Analysis of key functions, equipment and

Transcription

1 Executive Summary of the deliverable The COMBO project will propose and investigate new integrated approaches for Fixed Mobile Convergence (FMC) in broadband access and aggregation networks. COMBO targets an optimal and seamless quality of experience for the end user together with an optimized network infrastructure ensuring increased performance, reduced cost and reduced energy consumption. This D3.1 document is the first deliverable of Work Package 3 Fixed Mobile Convergent Architectures. As a main output of Task 3.1 Overall analysis and recommendations, this document prepares the overall work of the work package through a preliminary identification and analysis of network functions, equipment and infrastructures to be implemented in fixed/mobile convergence scenarios. Together with outputs from WP2 Framework definition, Architecture and Evolution, this document will serve as a basis for the definition and analysis of FMC network scenarios to be developed in WP3. The architecture and roles of key equipment in today s networks are first described with a structural viewpoint, considering the various network types covered by COMBO (fixed, mobile, Wi-Fi). This description includes the main relations between key equipment as well as some protocol stack examples illustrating the actual complexity of current networks. A global picture, called current FMC network, is also described similarly, so as to show how current networks already incorporate some degree of convergence between fixed, Wi-Fi and mobile. So as to prepare future functional analyses of FMC network scenarios, the main network functions implemented in current networks are listed and classified according to eleven functional groups. Starting from the current network descriptions, an analysis of differences between fixed, Wi- Fi and mobile networks is then undertaken. This comparative analysis is performed both with a structural viewpoint (considering the various network segments) and with a functional viewpoint (based on the functional groups defined previously). It is followed by an identification of gaps towards convergence, based on an analysis of the technical problems which have to be solved to answer to use cases derived by WP2. The impact of new frameworks (Cloud Computing, Software Defined Networking, Network Function Virtualization, Autonomic Networking) is then described through a top down approach of network convergence, analyzing the potential benefits of such emerging industry frameworks for network convergence. No final report - Waiting for acceptance from the European Commission Page 1 of 119

2 Finally, the activities undertaken up to now have put to the forefront the following technological concepts, which should be key enablers for network convergence: an Unified optical access & aggregation network; Heterogeneous radio access networks; BBU hostelling and mobile fronthaul technologies; Generalized handover and advanced offloading. No final report - Waiting for acceptance from the European Commission Page 2 of 119

3 Table of content EXECUTIVE SUMMARY OF THE DELIVERABLE 1 TABLE OF CONTENT 3 1 INTRODUCTION 5 2 IDENTIFICATION OF KEY EQUIPMENT AND FUNCTIONS IN TODAY S NETWORKS Description of today s networks and roles of key equipment Description of key functional groups and functions in today s networks 19 3 COMPARATIVE ANALYSIS OF FIXED, WI-FI AND MOBILE NETWORKS AND GAPS TOWARDS CONVERGENCE Comparative analysis of fixed, Wi-Fi and mobile networks Identification of gaps towards FMC networks 45 4 IMPACT OF NEW FRAMEWORKS Cloud Computing Software Defined Networking Network Function Virtualization Autonomic Networking 59 5 SOME TECHNOLOGICAL ENABLERS FOR CONVERGENCE Unified optical access & aggregation network Heterogeneous radio access networks BBU hostelling and mobile fronthaul technologies Generalized handover and advanced offloading 71 6 REFERENCES 75 7 CONCLUSION 80 8 GLOSSARY 111 No final report - Waiting for acceptance from the European Commission Page 3 of 119

4 9 LIST OF TABLES LIST OF FIGURES LIST OF AUTHORS List of reviewers Approval Document history Distribution list FURTHER INFORMATION 119 No final report - Waiting for acceptance from the European Commission Page 4 of 119

5 1 INTRODUCTION This D3.1 document is the first deliverable of Work Package 3 Fixed Mobile Convergent Architectures. As a main output of Task 3.1 Overall analysis and recommendations, this document prepares the overall work of the work package through a preliminary identification and analysis of network functions, equipment and infrastructures to be implemented in fixed/mobile convergence scenarios. Together with outputs from WP2 Framework definition, Architecture and Evolution, this document will serve as a basis for the definition and analysis of FMC network scenarios to be developed in WP3. Section 2 describes first the architecture and roles of key equipment in today s networks with a structural viewpoint, considering the various network types covered by COMBO (fixed, mobile, Wi-Fi). This section also lists and classifies the main network functions implemented in networks according to eleven functional groups. Section 3 then analyses the differences between fixed, Wi-Fi and mobile networks. This comparative analysis is performed both with a structural viewpoint (considering the various network segments) and with a functional viewpoint (based on the functional groups defined previously). It is followed by an identification of gaps towards convergence, based on an analysis of the technical problems which have to be solved to answer to use cases derived by WP2. Section 4 focuses on the impact of new frameworks (Cloud Computing, Software Defined Networking, Network Function Virtualization, Autonomic Networking) through a top down approach of network convergence, analyzing the potential benefits of such emerging industry frameworks for network convergence. Finally, whereas section 4 concentrates on architectural aspects, section 5 derives some technological enablers for network convergence, which are believed to be keys to future FMC networks. Appendix A.1 presents in more details the functions and functional groups elaborated on in section 2. A more detailed description of mobile network architecture is also given in appendix A.2. No final report - Waiting for acceptance from the European Commission Page 5 of 119

6 2 IDENTIFICATION OF KEY EQUIPMENT AND FUNCTIONS IN TODAY S NETWORKS This section identifies which equipment hosts which functions. It describes first the architecture and roles of key equipment in today s networks with a structural viewpoint, considering the various network types covered by COMBO (fixed, mobile, Wi-Fi). The framework reference used for these descriptions was developed in COMBO deliverable D2.1 [1]. This section also lists and classifies the main network functions implemented in networks according to eleven functional groups. 2.1 Description of today s networks and roles of key equipment This section presents structural descriptions of today s networks, for the three network types covered by the project: fixed, mobile and Wi-Fi. This analysis is followed by an overall structural description of the FMC features of current networks, so as to show the degree of convergence of existing networks Fixed networks Figure 1: Fixed Network Architecture Reference Description: The above figure represents six cases, two are Digital Subscriber Line (DSL) access using legacy copper access to the users, and three are Fibre accesses (FTTx) using No final report - Waiting for acceptance from the European Commission Page 6 of 119

7 new fibre infrastructure to the users and a last case (cable access) relies on a Hybrid Fibre Coaxial (HFC) network. The reader interested in a more detailed description of equipment in fixed networks can refer to reference [3]. The five first cases rely on the same infrastructure for the aggregation network and the interconnection to the ISP. Traffic flows from all users attached to a given Central Office (CO) are L2 multiplexed and are forwarded towards the BRAS. L2 multiplexing was originally performed in ATM, whereas Ethernet multiplexing is now more common. Ethernet aggregation could be achieved natively, with the use of VLAN/QinQ methods, or via an MPLS aggregation network, supporting Ethernet over MPLS. The Broadband Remote Access Server (BRAS) performs either L2 switching/bridging or L3 routing before directing the flows towards the edge router of the ISP to which the user is connected. It provides layer 2 connectivity through either transparent bridging or PPP sessions over Ethernet or ATM sessions and it provides layer 3 connectivity and routes IP traffic through an Internet service provider s backbone network to the Internet. AAA servers accessed through the edge router are responsible for both Access Control to the ISP and Address Allocation. When the ISP also controls access and aggregation (i.e. the ISP operates the Access Network), the BRAS and Edge Router can be merged into a single equipment. The CO may implement other security functions such as mapping between a L2 (ATM or Ethernet) address and a user s identity. The first DSL access architecture relies on DSL Access Multiplexers (DSLAMs) located at the CO. Each user has a dedicated point-to-point DSL connection carried on the copper loop linking the user and the CO. ATM is used between the DSL modem located in the Residential Gateway (RGW) and the DSLAM, transparently to the user. The second DSL architecture has been designed to make the DSLAM closer to users in order to enhance the upstream and downstream rates offered to the user. The DSLAM is now located at the Cabinet, which implies that the Cabinet now supports active equipment. As previously, ATM is used between the DSL modem in the RGW and the DSLAM, and the Cabinet is linked to the CO with an optical fibre. The first fibre access is Fibre to the Home (FTTH), in which a modem with optical interfaces replaces the DSL modem in the RGW. This architecture relies on a Pointto-Point fibre between the user and the CO. This architecture requires an interface per user at the CO, as the previous DSL based architectures. The second fibre access is also FTTH, but limits both the number of fibres and the number of interfaces at the CO by relying on a Point-to-Multipoint distribution tree. The optical modem includes an Optical Network Termination (ONT), dedicated to a single user. Multiple users (typically up to 64 or 128 users) can share a single fibre that is used for implementing a Passive Optical Network (PON). The Access Node (AN) in this architecture is an Optical Line Termination (OLT), which is located at the CO. Physical, passive multiplexing is performed in the upstream direction at one or several splitters between the ONTs and the OLT. Downstream traffic is copied on all fibres at the splitter, and filtered by each ONT. Traffic from multiple OLTs are L2 multiplexed in the CO in which multiple OLTs are connected. No final report - Waiting for acceptance from the European Commission Page 7 of 119

8 The third fibre access is Fibre to the Building (FTTB). In this case, the optical modem is an Optical Network Unit (ONU), in the building; it is part of the access network. Each user is linked over Ethernet to the ONU. The RGW now requires a simple Ethernet interface. A first multiplexing stage is thus provided within the building. The rest of the architecture is similar to the previous one. The cable access shown at the bottom of the figure is currently proposed by Cable Operators who implement DOCSIS. Usually, Cable Operators are not originally Telecom Operators and therefore do not rely on the same aggregation network as the one applicable to the other five cases. The RGW now contains a Cable Modem (CM) which links the user to a fibre node by a coaxial cable shared by multiple users (typically up to 500) thanks to multiple RF amplifiers. At the Fibre Node (FN) the traffics from several cables are multiplexed and transmitted over fibre to the Cable Modem Termination System, which is the interface to the IP backbone. The CMTS may serve typically users, and is equivalent to a CO. Protocol stack example: Figure 2: Example of Protocol Stack: user plane in a Fixed Network A possible protocol stack for the first DSL access is represented in the above figure. Some protocols are passed transparently in some equipment (for example, the DSLAM does not access the IP layer); therefore, the above figure shows protocol layers only where the protocol is accessed by the equipment. No final report - Waiting for acceptance from the European Commission Page 8 of 119

9 The above figure assumes that the ISP and the access provider are different actors. The ISP relies on the AP to carry user flows between the LER and the RGW. This is done thanks to different tunnelling techniques: PPP (Point to Point Protocol) is used between the RGW and the LER, as the ISP authenticates the user, and allocates an IP address in the ISP controlled address block; PPPoE is used between the RGW and the BRAS within the Ethernet aggregation network; L2TP is used between the BRAS and the LER in order to transparently carry the user s IP traffic in the IP part of the aggregation network. If a single actor plays both roles (access provider and ISP), the LER can be merged with the BRAS and the L2TP layer is redundant Mobile networks Home Building Cabinet Central Office (CO) Main CO Core CO Access Aggregation Mobile Backhaul MASG Metro Access Router Core Network Mobile Core MME SGSN Internet Mobile 2G/3G LTE/4G UE BTS Node B enode B CSG MW link nodes Copper links Aggr. Switch 1 CWDM optionally Aggr. Switch 2 RNC BSC IP-Sec GW epdg SGW PGW GGSN HSS IP Backbone LER AAA servers PCRF Figure 3: Mobile Network Architecture Reference Description: The reference network architecture for the currently deployed mobile network is presented in Figure 3. According to the linear network section reference indicated in the top of the figure, the main elements and respective position for the 3GPP based radio access network, mobile backhaul and core network are indicated. In the radio access network (blue region), it is indicated the wireless connection between a User Equipment UE and radio base stations BTS, NodeB and enodeb, which make use of radio access technologies from the different generations 2G, 3G and LTE/4G respectively. The macro site antennas representation along the access network indicates the existence of macro sites with mobile backhaul links being No final report - Waiting for acceptance from the European Commission Page 9 of 119

10 connected in the network level of cabinets or even directly to the COs. The first mobile backhaul links connecting the Cell Sites Gateways CSG, at the boundary of the RAN and the backhaul network (yellow region) can make use of different transport physical media, such as fibre, copper or micro wave links, according to the available network infrastructure. Note that mobile base-stations need a highly accurate timing signal that has to be shared across the network. If an individual base-station drifts outside of the specified +/-50 PPB (Parts per Billion) limit, mobile handoff performance degrades resulting in high disconnected calls rate and poor data services quality. More details on synchronization issues can be found in appendix A.1.6. At the aggregation level, the network architecture for mobile system is similar to the fixed network and can share the same infrastructure (this practice has been observed in limited deployments), typically with fibre based transport, e.g. CWDM. In this level of the network, the deployment of a chain of aggregation switches can occur until the Mobile Aggregation Site Gateway MASG is reached. In Figure 3, some of the core elements, such as Base Station Controller BSC and Radio Network Controller RNC, are partially placed in aggregation network region since they can be found distributed in the aggregation sites near the core. The 2G/3G radio access network connection from the BTS/NodeB equipments goes through the MASG to the BSC/RNC, respectively. In contrast, the connection between the LTE/4G enodeb and the SGW/PGW node can be done directly from one of the aggregation switches. For more details about the 3GPP mobile network architecture regarding control plane, data plane and interface connections from radio access to core elements please see Appendix A.3 and [4][5][6][7][8]. In case of 4G/LTE the core network is called the Evolved Packet Core EPC within the LTE System Architecture Evolution SAE concept. In LTE all traffic is considered packet data, including voice. The Mobility Management Entity MME and Serving GPRS Support Node SGSN can be provided as separated physical nodes or as one common, fully integral, physical node, as here represented. Control signalling, for example for mobility, is handled by the MME, while the Serving GPRS Support Node SGSN deals with both control and user plane messages. The Serving Gateway SGW routes the user plane communication from the UE to the Packet Data Network Gateway PGW. The UE is attached to the same SGW during the complete session. The SGW contains functionalities such as local anchor for mobility, network routing information, charging for roaming users and lawful Interception. The SGW is only changed when the UE moves to a new SGW pool area while the PGW is normally kept as long as the UE is attached to the network. The PGW is the gateway between the internal EPS network and external IP Networks, for example, the Internet or a corporate LAN. The UE can be connected to several PGW nodes simultaneously to access multiple Packet Data Networks. The PGW contains functions such as QoS Policy Control and Enforcement, Packet Filtering and Charging. In Figure 3, the SGW and PGW are shown integrated in the same node, together with the Gateway GPRS Support Node GGSN, that provides access to external IP data packet networks (Internet). No final report - Waiting for acceptance from the European Commission Page 10 of 119

11 Among the 4G registers, there is the Home Subscriber Server HSS, which is the database that holds the subscription information for UE subscribing to the Evolved Packet Switch EPS network. The HSS stores, for example, the location of the UE (on MME node level), and authentication parameters. The HSS is an evolution of the Home Location Register HLR. For policy control decisions and flow based charging control functionalities, the Policy and Charging Rules Function PCRF core element is applied. The PCRF takes decisions based on subscription information. The functions performed by PCRF also include authorization of QoS resources, IP flow mobility routing rules and possibly monitoring control. The connections of devices using non 3GPP radio access technologies are assumed trusted or untrusted, according to the operators criteria. Untrusted non-3gpp accesses interwork with the EPC via a network entity called the Evolved Packet Data Gateway epdg. The main role of the epdg is to provide security mechanisms such as IPsec tunnelling of connections with the UE over an untrusted non-3gpp access. Non-3GPP accesses considered as trusted can interact directly with the evolved packet core elements. Also, the AAA server node is responsible to interface non 3GPP access to HSS, providing e.g. access subscription data or any other information available in the HSS. Protocol stacks examples: Figure 4: Radio Protocol - Control Plane [5]. No final report - Waiting for acceptance from the European Commission Page 11 of 119

12 Figure 5: Radio Protocol - User Plane [5]. The radio protocol architecture for control plane and data plane is shown in Figure 4and Figure 5 respectively. S1-MME Reference point for the control plane protocol between LTE/eNode Band MME and also the transport for NAS signalling (between UE and MME). S1-U interface provides transport of user data traffic received from the terminal between the enodeb and the Serving GW. It is a reference point between LTE/eNodeB and Serving GW for the per bearer user plane tunnelling and inter enodeb path switching during handover. In addition, the UE user data is encapsulated in GTPv1-U and sent to the Serving GW. S5/S8 interface provides user plane tunnelling and tunnel management between Serving GW and PDN GW. It is used for Serving GW relocation due to UE mobility and if the Serving GW needs to connect to a remote PDN GW for the required PDN connectivity. S8 has the same function as S5 but is used for inter-plmn connectivity, and includes a border GW function for the inter- PLMN connectivity. S11 acts as a reference point between MME and Serving GW. As user plane and control plane are separated, MME needs to communicate with Serving GW in order to activate/manage UE sessions in the user plane. S11 interface keeps the control and user plane procedures in sync for a terminal during the period that the terminal is seen active/attached in the EPS. SGi is the reference point between the PDN GW and the packet data network. Packet data network may be an operator external public or private packet data network or an intra operator packet data network, e.g. for provision of IMS services. This reference point corresponds to Gi for 3GPP accesses. No final report - Waiting for acceptance from the European Commission Page 12 of 119

13 2.1.3 Wi-Fi networks Figure 6: Wi-Fi Network Architecture Reference Description: Wireless access allows cheaper deployment of local area networks (LANs). Also spaces where cables cannot be run, such as outdoor areas and historical buildings, can host wireless LANs. Nowadays many Wireless LANs are being deployed at airports, business class hotels and other public areas. With Wireless LANs, applications such as web browsing, ing and file transferring can now be used at public places in the same way as in the office or in the home. IEEE x wireless networks, known as Wi-Fi, are the most common access technology among Wireless LANs and especially their popularity is increased with the spread of smart phones. Today Wi-Fi networks can be adopted to several network areas, for example they can be seen as a complement to Mobile networks and also they can be used as an extension to fixed network coverage area by providing mobility to users. Four use cases can be defined for today s Wi-Fi network deployments: Public Wi-Fi deployment: In public areas like airports, streets, malls etc., ISPs deploy APs (Access Points) to provide internet access to customers. The coverage area of one or more (interconnected) Wi-Fi AP are distributed along a site are called Hotspot. Using their Wi-Fi end devices, users connects to a Public AP and then after an authentication process they access to Internet. Authentication process is done using Portal and Radius server and then the Access Router routes authorized user traffic to packet data networks. All APs operating on a hotspot point are managed by an AP Controller. Fixed Access/Aggregation network connects these Public APs to the Wi-Fi No final report - Waiting for acceptance from the European Commission Page 13 of 119

14 core. Handover between APs is possible if multiple Wi-Fi access point topologies are deployed. Figure 7: Public Wi-Fi Deployment Residential Wi-Fi deployment: Using a Wi-Fi device, a residential user establishes Wi-Fi connections to an indoor Wi-Fi access point. The indoor AP is connected to ISP s Wi-Fi core through a Fixed Access/Aggregation network. The Residential Gateway authenticates Wi-Fi users using a pre-shared key (WEP, WPA ). Figure 8: Residential Wi-Fi deployment Community Wi-Fi deployment: The creation of Wi-Fi networks using CPEs (Customer Premise Equipment, for example a Residential Gateway) is quite recent. It is the same service that is offered in public places but from equipment located at homes. No final report - Waiting for acceptance from the European Commission Page 14 of 119

15 Figure 9: Community Wi-Fi Deployment Operator managed business Wi-Fi deployment: Operators deploy private Wi-Fi networks to business premises like hotels to enable customers from the corresponding business unit access internet. The management of the network is performed by the operator and internet access can be either free or charged based on the business units policy. Figure 10: Operator Managed Business Wi-Fi Deployment No final report - Waiting for acceptance from the European Commission Page 15 of 119

16 Protocol stacks examples: Figure 11: Protocol Stack Example for a Wi-Fi Network (Data Plane) Figure 12: Protocol Stack Example for a Wi-Fi Network (Control Plane, Public Deployment) No final report - Waiting for acceptance from the European Commission Page 16 of 119

17 Figure 13: Protocol Stack Example for a Wi-Fi Network (Control Plane, Residential/Community Deployment) FMC networks Figure 14: Example of FMC in current Network No final report - Waiting for acceptance from the European Commission Page 17 of 119

18 Description: Figure 14 shows the state-of-the-art of fixed mobile convergence in today s access and aggregation networks. In principal, a certain degree of structural convergence is already realized in today s access and aggregation network deployments as shown in Figure 14. With the evolution towards All-IP, mobile base stations and fixed access nodes (DSLAM, OLT) have been connected to a common Ethernet based aggregation network that is more suitable for IP packet transport than the legacy transmission technologies (SDH/ATM). In addition, backhauling of the different base stations and access nodes are often realized via CWDM technology in order to minimize the amount of fibre in the aggregation network section. Fixed, mobile and Wi-Fi networks are structural converged in the aggregation network section by sharing the same fibre infrastructure and equipment. A higher level of structural convergence already exists for the Wi-Fi deployment, since the Wi-Fi access points are typically connected by the fixed DSL access network via ADSL/ADSL2+ and VDSL2. In contrast to that backhaul of mobile base stations and fixed access nodes (DSLAM) in the access part is provided by individual fibres, but the same technology (CWDM) is used. FTTH residential users are served by a dedicated technology (GPON) and fibres in access. So in the access there is limited structural convergence today. All access and aggregation networks (fixed, mobile, Wi-Fi) are usually connected to a common IP backbone network that is based on MPLS technology. The control and user plane of the different access platforms are mostly separated, especially between the fixed and the mobile network. In the Wi-Fi case some central components of the fixed and/or mobile network might be used, e.g. AAA information Protocol stack example UE enode B CSG Eth. Switch 1 Eth. Switch 2 Ethernet Aggregation MASG Access Router LER IP Backbone LER EPC S-GW IP PDPC RLC MAC L1 Relay PDPC GTPv1-U UDP RLC IP MAC Eth.MAC L1 Eth. Eth MAC Eth. Eth WDM* Eth MAC Eth. Eth. WDM* Eth MAC Eth Eth. IP Eth MAC Eth MAC Eth. Eth. IP MPLS Eth MAC Eth MAC Eth. Eth. WDM IP MPLS Eth MAC Eth.MAC Eth. Eth. WDM GTPv1-U UDP IP Eth.MAC Eth. * optionally Figure 15: LTE protocol stack including Ethernet aggregation network No final report - Waiting for acceptance from the European Commission Page 18 of 119

19 2.2 Description of key functional groups and functions in today s networks This section describes all key functions of current networks, organized according to eleven functional groups which were defined by COMBO project. These functional groups represent key sets of basic functions which are the enablers of high-level functionalities in the network. An important point is that a given function or functional group is in general not implemented in a single equipment, and can be distributed on several equipment. The distribution of functions in equipment is the objective of the functional analysis. Part 3 of this document will in particular propose directions for improvement of this distribution, taking into account use cases described in WP2 D2.1 [1]. This section presents an overall description of each functional group and some tables listing the main functions included in the functional group. The reader interested in the descriptions of functions themselves can find more details in appendix A.1. The functions mentioned in the tables below are also put in one or several of the three following categories: Control plane, Data plane or Management plane. They are the three basic components of a telecommunication network architecture. The control plane is the part of a network that carries signalling traffic and is responsible for routing rules. Control packets originate from or are destined for a router. Functions of the control plane include system configuration and management. The control plane and management plane serve the data plane a.k.a. forwarding plane or user plane, which bears the user traffic. The management plane, which carries administrative traffic, is sometimes considered as a subset of the control plane. Note that for some functions, several OSI layers are implicated and a very detailed description of the functions would be needed for clarifying the roles of a given function in a given layer. Thus, all OSI layers given in the following tables have to be considered as an indication of the typical scope of each function Forwarding In this context forwarding means the procedure of the data transmission through a network section or element from an ingress interface to an egress interface. Name of the function OSI layer Plane L2 Forwarding 2 Data plane Ethernet switching 2 Data plane MAC address learning 2 Data plane/control No final report - Waiting for acceptance from the European Commission Page 19 of 119

20 VLAN handling 2 Data plane IEEE 802.1AD Provider Bridges (Q-in-Q) 2 Data plane IEEE 802.1ah Provider Backbone Bridging (MAC-in- MAC) 2 Data plane L3 Forwarding 3 Data plane IP Routing 3 Data/Control plane IPv6 support 3 Data/Control plane Network Address Translation (NAT) 3 Data/Control plane Multicast 2-3 Data/Control plane L3 Multicast 3 Data/Control plane Internet Group Management Protocol (IGMP) 3 Control plane L2 Multicast 2 Data/Control plane IGMP snooping 2/3 Data/Control plane Multiprotocol Label Switching (MPLS) 2/3 Data/Control plane PE MPLS (Multiprotocol Label Switching) 2/3 Data/Control plane Seamless MPLS 2/3 Data/Control plane Point-to-Point Protocol (PPP) 2 Data plane Layer-2-Tunnelling-Protocol (L2TP) 2 Data plane Generic Routing Encapsulation (GRE) 3 Data plane Proxy Mobile IP (PMIP) 3 Data/Control plane GPRS Tunneling Protocol (GTP): 3-7 Data/Control plane GTP control plane (GTP-C) 5-7 Control plane GTP user plane (GTP-U) 3-7 Data plane Traffic Offloading: mobile traffic through WLAN 2-3 Data/Control plane HTTP traffic redirect 3-7 Data plane Table 1: Relevant Forwarding functions in access and aggregation networks No final report - Waiting for acceptance from the European Commission Page 20 of 119

21 2.2.2 Automatic configuration and management This functional group includes the functionalities related to configuration and management that does not need any manual operation. Name of the function OSI layer Plane Physical medium resources management 1 Control plane Radio resources management 1 Control plane Device management 1-7 Control/Management Plane Logical Link management (VLANs ) 2 Management Plane IP infrastructure management 3 Management Plane Customer IP pools management 3 Management Plane Subscriber database (AAA, HLR, HSS) 1-7 Control/Management Plane SNMP 7 Control/Management Plane Table 2: Relevant functions for Automatic Configuration and Management Resilience This group includes the functions related to network and service resilience. Resilience is the ability of a service or a network of recovering itself in a quick and easy way after being disturbed, maintaining an acceptable level of service when faults occur in comparison to normal operation. The main functionalities included in this functional group are redundancy and load balancing. Reminder: note that for some functions, several OSI layers are implicated and a more detailed description of the function would be needed for clarifying the roles at each layer. Name of the function OSI layer Plane Redundancy 1-7 Data/Control Plane Performance monitoring 1-7 Data/Control Plane Fault detection 1-7 Data/Control Plane No final report - Waiting for acceptance from the European Commission Page 21 of 119

22 Network switching 1-7 Data/Control Plane Trouble reporting 1-7 Control Plane Normal service restoration 1-7 Data/Control Plane Load balancing 3-7 Data/Control Plane Queue management 4-7 Data Plane Performance monitoring 4-7 Data/Control Plane Forwarding requests 4-7 Data/Control Plane Security point (e.g.: against DDoS) 3-7 Data/Control Plane Session persistence (multiple requests from the same user) 7 Data/Control Plane Caching 7 Data/Control Plane Table 3: Relevant Resilience functions This functional group has many relationships with other functional groups; however, the most related functional groups are the following ones: OAM & Management, where network monitoring and fault detection mechanisms are included. Policy and Charging, where the quality of service is established and enforced. Resilience functionalities can be found in all network segments, such as in home/business networks (e.g.: multi-homing through multiple ISP or through multiple links), access networks (e.g.: PON protection), aggregation networks (e.g.: link aggregation IEEE 802.1ax) and transport networks (e.g.: MPLS redundancy), and in different network elements, such as CPEs, access network nodes, switches, routers, services specific devices, etc. and in network protocols (RSTP, RSVP, MPLS Fast ReRoute, etc.) Security This group includes the functions related to network security. These can be roughly divided into two groups. The first includes the functionalities for the protection of communications either between the user and the Access Server or between the operator s network nodes. It deals with the following threats: Unauthorized access to user ID and connection set-up information (spoofing and theft of service); No final report - Waiting for acceptance from the European Commission Page 22 of 119

23 Interception of user data flows (threat to data confidentiality); Unauthorized modification of user data flows (threat to data integrity); The second group includes functionalities aiming at preventing and monitoring unauthorized access and misuse of the network resources. It deals with the following threats: Interception of user location (threat to user s privacy, especially in case of wireless access); Unauthorized access to network resources; Unauthorized use of a stolen device. Name of the function OSI layer Plane End-user Authentication 2-3 Control DHCP relay option Control PPP over Ethernet (used in RADIUS server transactions Session key agreement and cryptographic tunnel management 2-3 Control 2-3 Control Firewalls 3-7 Control Layer 2 security 2 Control IPSec 2-3 Data/Control 802.1x EAP 2-3 Control Dynamic ARP Inspection, ARP-Reply-Agent, MAC- Address Limitation, MAC Anti-Spoofing 2 Control Access Control 2-3 Data/Control Packet blocking, rate limitation 2-3 Data Service blocking 3-7 Data/Control Table 4: Relevant Security Functions No final report - Waiting for acceptance from the European Commission Page 23 of 119

24 2.2.5 OAM & Management Operations, Administration Maintenance (OAM) and management are the processes, activities, tools and standards involved with operating, administering, managing and maintaining a system. Note that this functional group is close to the Automatic Configuration and Management functional group, as both deal with the management of the network, we just made a distinction between what is done automatically and what is done manually. We describe the logical model by management tasks of FCAPS. For details please see appendix A.1.5. Name of the function OSI layer Plane Fault management 1-7 Physical/Network/Service/Business Reliability, Availability, Survivability (RAS) quality assurance 1-7 Physical/Network/Service Alarm surveillance 1-7 Physical/Network Fault localization 1-7 Physical/Network Fault correction 1-7 Physical/Network/Service/Business Testing 1-7 Physical/Network Trouble administration 1-7 Physical/Network/Service/Business Configuration management 1-7 Physical Hardware provisioning and inventory reporting Software provisioning and inventory reporting 1-7 Physical/Network 1-7 Physical/Network/Service Protection switching configuration 1-7 Physical/Network/Service Trail termination configuration 1-7 Physical/Network/Service Adaptation configuration 1-7 Physical/Network/Service Connection configuration 1-7 Physical/Network/Service Degrade thresholds setting 1-7 Physical/Network/Service Alarm severity provisioning 1-7 Physical/Network/Service No final report - Waiting for acceptance from the European Commission Page 24 of 119

25 Alarm reporting control provisioning 1-7 Physical/Network/Service Performance Monitoring thresholds setting 1-7 Physical/Network Tandem Connection Monitoring 1-7 Physical/Network Date and time parameters 1-7 Physical/Network Accounting management 1-7 Service/Business Policies 1-7 Physical/Network/Service/Business Service Level Agreement 1-7 Physical/Network/Service/Business Quality of Service 1-7 Physical/Network/Service/Business Quality of Experience 1-7 Service/Business Performance management 1-7 Physical/Network/Service/Business Quality assurance 1-7 Physical/Network/Service/Business Monitoring 1-7 Physical/Network/Service/Business Management control 1-7 Physical/Network/Service/Business Analysis 1-7 Physical/Network/Service/Business Security management 1-7 Physical/Network/Service/Business Synchronization Table 5: Relevant OAM & Management Functions This functional group includes the functionalities related to synchronization of frequency, phase and time in networks. This means the transport and alignment of these three units. The reader interested in more details about technological aspects of synchronization is referred to COMBO deliverable D2.2 [2]. Frequency Synchronization: Aligning clocks of the network elements with respect to frequency. Clocks may have different time but rates are same. No final report - Waiting for acceptance from the European Commission Page 25 of 119

26 Figure 16: Frequency Synchronization Phase Synchronization: Aligning clocks of the network elements with respect to phase. Clocks may have different time but rates and click instants are same. Figure 17: Phase Synchronization Time Synchronization: Aligning clocks of the network elements with respect to time. The two clocks must utilize the same epoch. Time synchronization implicitly includes phase and frequency synchronization. Clocks have same time of the day, rate and phase. Figure 18: Time Synchronization Name of the function OSI layer Plane Frequency Synchronization 1-7 Control Plane* Phase Synchronization 1-7 Control Plane* No final report - Waiting for acceptance from the European Commission Page 26 of 119

27 Time Synchronization 1-7 Control Plane* *: in-band signalling Table 6: Relevant Synchronization Functions Note that even if the traffic related to synchronization is in-band signalling, it is considered as Control Plane because it is not related to user data Policy and Charging This group includes the functions related to Policy, Charging and Enforcement performed on a network scope (see 3GPP related functions and processes) and policing functions performed on a node scope (see packet processing related functions, common to both fixed and mobile equipments). Detailed description of each function may be found in Appendix A.1.7. Policy and Charging related functions are: Name of the function OSI layer Plane Policy Control 2-7 Data/Control Plane Gate Enforcement 2-3 Data/Control Plane QoS enforcement 2-3 Data/Control Plane QoS definition/authorization 7 Data/Control Plane QoS rules 7 Data/Control Plane Charging Control 2-7 Data/Control Plane Policy & Charging Control decision 2-7 Data/Control Plane PCC architecture & functions Data/Control Plane PCRF - Policy and Charging Rules Function 3-7 Data/Control Plane PCEF - Policy and Charging Enforcement Function 3-7 Data/Control Plane AF - Application Function 3-7 Data/Control Plane SPR - Subscription Profile Repository 3-7 Data/Control Plane No final report - Waiting for acceptance from the European Commission Page 27 of 119

28 OFCS - Offline Charging System 3-7 Data/Control Plane OCS - Online Charging System 3-7 Data/Control Plane Network authorization 3-7 Data/Control Plane BBERF- Bearer Binding and Event Reporting Function Table 7: Relevant Policy and Charging Functions 3-7 Data/Control Plane The PCC architecture includes the following reference points: Rx: the Rx reference point resides between the AF and the PCRF. Gx: the Gx reference point resides between the PCEF and the PCRF. Sp: the Sp reference point lies between the SPR and the PCRF. Gy: the Gy reference point resides between the OCS and the PCEF. Gz: the Gz reference point resides between the PCEF and the OFCS. S9: The S9 reference point resides between a PCRF in the HPLMN (H-PCRF) and a PCRF in the VPLMN (V-PCRF). Packets policing related functions are: Name of the function OSI layer Plane Classification 2-3 Data/Control Plane Metering 2-3 Data/Control Plane Marking 2-3 Data/Control Plane Scheduling 2-3 Data/Control Plane Shaping 2-3 Data/Control Plane Table 8: Relevant Packet Policing Functions Subscriber data and session management In this group, all functions related to the management of the subscriber are addressed: subscriber data management (HLR, HSS, AAA like Radius server) and subscriber session management. Name of the function OSI layer Plane No final report - Waiting for acceptance from the European Commission Page 28 of 119

29 Fixed network Session creation and detection (PPP, DHCP, IP...) 2-3 Data and Control Plane Circuit ID and Remote ID insertion 2-3 Control Plane Session Identification 2-3 Control Plane Session Authentication 2-3 Control Plane Session Authorization 2-3 Control Plane Resource Admission Control 2-3 Control Plane IP configuration 3 Data and Control Plane Session accounting 2-3 Data and Control Plane Session monitoring 2-3 Data and Control Plane Session modification 2-3 Data and Control Plane Session termination 2-3 Data and Control Plane AAA client 2-3 Control Plane Session database 2-3 Control Plane Mobile network Attach Request 3 Control plane Identification, Authentication, and Security 3 Control plane Location Update 3 Control plane Session Creation 3 Control plane Attach Accept 3 Control plane Session Reconfiguration 3 Control plane Table 9: Relevant Subscriber Data and Session Management Functions No final report - Waiting for acceptance from the European Commission Page 29 of 119

30 2.2.9 Mobility This group includes the mobility of the device/user and the nomadism (access to the network from different locations). Mobility is the ability to support from low to high speed users, providing a seamless service experience and session continuity. Name of the function OSI layer Plane Mobility Client 7 Control plane Mobility Anchor 3 Data plane Mobility Decision 3 Control plane Measurement and reporting 2 Control plane Paging 3 Control plane Localization 3 Control plane Table 10: Relevant Mobility Functions Legal interception and data retention This group includes the main requirements from authorities for legal interception of phone calls, SMS or data sessions, mirroring and the needs of access to subscriber activities such as logs of calls, visited websites or services. Name of the function OSI layer Plane Legal interception 1-7 Data/Control Access line identification (Identification of the access line of the end user to be intercepted) 1-2 Data/Control Content decryption (if it is encrypted) 2-7 Data Content duplicating 2-3 Data Port Mirror 2-3 Data Content encryption with a standard key (optionally) 2-7 Data Content forwarding to law enforcement agency 2-3 Data/Control Emergency calls 1-7 Data/Control Access line identification (Identification of the 1-2 Data/Control No final report - Waiting for acceptance from the European Commission Page 30 of 119

31 physical location of the calling end) Priority for emergency calls 2-7 Data Route emergency calls to recorded multilingual announcements 3-7 Data/Control Table 11: Relevant Legal Interception and Data Retention Functions Traffic analysis Traffic monitoring and analysis is the set of functions that can be implemented by a network operator to assess the usage of its resources by traffic flows in order to implement various policies including: Derive usage metrics e.g. for computing traffic matrices, or to preventively increase resources in order to avoid congestion; more generally, derive usage statistics. Detect unexpected traffic variations that could correspond to an attack such as a Distributed Denial of Service (DDOS), and block selectively traffic in order to avoid congestion Detect unauthorized traffic flows, which can be illegal (e.g. child pornography) or can be excluded from a SLA (e.g. P2P traffic). Name of the function OSI layer Plane In-service OAM Performance Monitoring 2-3 Data/Control Inserting, extracting and analyzing in band OAM flows 2-3 Data/Control Out-of-band Performance Monitoring 3 Data/Control With stand alone test equipment 2-3 Data With local tools 3 Data Traffic metering 2 Data/Control With equipment counters and e.g. SNMP 2 Data/Control Packet flow analysis 2-3 Data/Control With local tools (.g. NETFLOW) 3 Data Deep Packet Inspection on captured traffic 2-6 Data Table 12: Relevant Traffic Analysis Functions No final report - Waiting for acceptance from the European Commission Page 31 of 119

32 No final report - Waiting for acceptance from the European Commission Page 32 of 119

33 3 COMPARATIVE ANALYSIS OF FIXED, WI-FI AND MOBILE NETWORKS AND GAPS TOWARDS CONVERGENCE After the description of the three network types and of the functional characteristics of current networks made in section 2, this section makes a comparison between these three network types (comparative analysis). This analysis of differences between fixed, Wi-Fi and mobile (section 3.1) allows then to identify gaps towards convergence (section 3.2), based on the use cases defined by WP2 in D2.1 [1]. 3.1 Comparative analysis of fixed, Wi-Fi and mobile networks This section stresses the main differences between the three network types (fixed, mobile, Wi-Fi) along various viewpoints (distribution of network elements, functional aspects). The analysis is first performed in each network segment and then according to the functional groups defined in section Home & Access The most obvious differences between fixed, Wi-Fi and mobile network can be found in the home and access network segment, especially due to the various characteristics of the different transmission mediums and technologies typically used in that network section. In contrast to wireless access, the fixed access operates over relatively stable transmission channels which make it much easier to provide guaranteed quality of service and high availability. Conversely, availability and service quality of wireless access is an issue, which gets worse considering mobility. For Wi-Fi the situation depends on the deployment, if it is a public or private (residential) setup. In the case of a public deployment the situation is more like the case of mobile access, while in a residential deployment it is most often part of the home network and therefore should be compared with the fixed access. Network termination: In the fixed access the network termination (NT) is part of a dedicated network element (e.g. DSL modem/router, GPON ONT, personal home router) that is separated from the customer premises equipment (e.g. PC, set-top box, phone).the NT forms the boundary to the provider network and is controlled by the network provider. It is the anchor point of many functions, e.g. OAM & management and automatic configuration. Often the NT is part of the Residential Gateway (RGW) equipment that also host functionality for setting up the home network, such as DHCP, NAT and Wi-Fi. The network provider is typically not able to control the network section behind the NT (i.e. the link between home network equipment and the NT) which makes it difficult to implement some functions on an end-to-end level, e.g. end-to-end QoS. An exception from this is the set-top box where the service provider has some extent of control of the equipment. In the mobile network the user equipment (UE) typically includes the network termination except for scenarios such as tethering (e.g. mobile terminals operating as Wi-Fi hotspot) and the fixed access emulation using residential gateways with 3GPP connectivity as shown in Figure 14. The NT in Wi-Fi networks depends on the deployment. The NT location of public Wi- No final report - Waiting for acceptance from the European Commission Page 33 of 119

34 Fi deployments is quite similar to the mobile network, whereas the residential Wi-Fi deployment can be compared with the fixed access, since NT is located at or before the Wi-Fi access point. Session Authentication: Session authentication to verify the identity of the users is realised in different ways in fixed, mobile and Wi-Fi networks. In the mobile network the credentials for the session authentication are typically based on identifiers stored on a SIM card that is part of the user equipment. In fixed and Wi-Fi networks typically a login/password combination is used, that must be configured manually by the end-user. Security (Encryption): In the mobile network, data encryption is performed on the access link by default due to the shared media characteristic. The security functions (including management of cryptographic keys) allow protecting the air medium between terminal and base station. The procedures to be applied are outlined in [9]. Due to the induced change of base stations by the mobility, the overall security control is performed between terminal and the mobile core (MME and home subscriber server (HSS) in case of LTE). The traffic running over the backhaul network is typically secured by IPsec. In the fixed access, data encryption is typically not performed in residential point-topoint deployments (e.g. DSL, Optical Ethernet). However, PON deployments work with encryption on the PON MAC layer to cope with the point-to-multipoint nature of that technology. In Wi-Fi networks it depends on the deployment. The Wi-Fi transmission over an open network is often not encrypted, whereas the traffic is encrypted in closed networks using encryption methods like WPA. Protocol layers: From a protocol layer point of view, the packet data transport is relatively flat in the fixed access and Wi-Fi compared with the mobile network as shown by the protocol stacks in Figure 2, Figure 5 and Figure 11. In fixed access deployments there are two leading concepts: PPPoE that enables a tunnelling of the user IP data via point-to-point protocol (PPP) over Ethernet on layer-2 IP over Ethernet (IPoE) that allows a direct transmission of IP packets over Ethernet using DHCP for IP auto-configuration On the contrary, in the mobile network, the user s IP data (layer-3) is encapsulated by layer-4 protocols (e.g. GTP) and transported in dedicated IP packets. Handling of IP addresses for the user is also very different between fixed and mobile: The fixed access network often handles one public IP address and several private addresses used in the home network. The RGW, at the boundary between home and fixed access, is thus implementing address translation (NAT function). Conversely, mobile terminals are connected to GGSN or P-GW in the mobile core network with private IP addresses, and NAT function is thus implemented deeper in the mobile core. No final report - Waiting for acceptance from the European Commission Page 34 of 119

35 In addition to the functional differences between fixed and mobile networks in the home and access segment, there are also significant structural differences. On the one hand, Wi-Fi access can be seen as a wireless extension of the fixed network, leading thus to mutualisation of fixed and Wi-Fi in this access segment. On the other hand, mobile access networks were built separately, not as an extension of fixed networks, but as dedicated access networks. This aspect is strongly related to the history of mobile communications and the current business models and relations between fixed and mobile operators. It has also some technical reasons, as some antenna sites (e.g. in rural areas) cannot be directly and easily connected to the existing fixed access networks. For all these reasons, parts of mobile backhaul networks consist today of dedicated fibre, copper, or micro wave point-to-point links used only for mobile backhauling. With the densification of cellular networks, leading to very large numbers of cells, using such dedicated links to connect the cells to the rest of the network might be a significant issue, both from an economical and technical viewpoint. Thus, the structural differences between fixed and Wi-Fi on the one hand, and mobile on the other hand, will have to be alleviated in the future, by targeting more mutualisation between mobile backhaul and fixed access Aggregation Regarding transport mechanisms, aggregation of segments of current fixed networks typically cover distances of a few tens or hundreds of kilometers using SONET/SDH, Ethernet and MPLS. In mobile networks, on the other hand, transport mechanisms in aggregation networks are migrating from voice-centric transport mechanisms, such as TDM and SDH transport, to adaptive and cost-effective IP and Ethernet transport mechanisms. For fixed network data plane, if Ethernet and MPLS are used in deployment of aggregation part, this aggregation segment is divided into Ethernet level (Ethernet over MPLS) and IP (IP over MPLS) level by the BRAS element that for instance establishes PPPoE with access element at Ethernet part and establishes L2TP with LER at the IP part. Ethernet level consists of switches performing generally up to two levels of switching. Here aggregation of subscriber traffic into different service providers is performed by Layer 2 traffic aggregation mechanisms such as PPP/PPPoE termination, PPP over L2TP, IP bridging (RFC2684/1483), ATM PVCs, IP VPNs, and VLANs. In case of mobile networks, there is a recent paradigm shift in implementing the aggregation networks. In 2G networks the GERAN radio access network contains the Base Transmitter Stations (BTS). Their traffic is aggregated by the Base Station Controllers (BSC) and forwarded to the core network: to the MSS for voice, and SGSN for packet data. In 3G network the situation is analogous: The NodeBs in the UTRAN are connected to the Radio Network Controllers (RNC) in the aggregation networks. RNCs are connected to the MSS/SGSN for voice/data forwarding. In 4G/LTE network architecture the design consideration was to make the network as flat and simple as possible. This resulted in the complete elimination of the aggregation functions: The enodebs in the eutran access network are directly No final report - Waiting for acceptance from the European Commission Page 35 of 119

36 connected to the MME and SGW with their control and user data interfaces, respectively. Of course, L2/L3 switches and routers used in the conventional fixed IP aggregation networks are still in operation for the 4G networks as well. Fixed networks constitute the aggregation segment of Wi-Fi access networks. Wi-Fi traffic is seen as arbitrary service provider traffic at this part of network and the network transports this traffic to the Wi-Fi core by using configured VLANs or VPNs. This network segment can differ in control and data plane when the AP Controller is deployed in this part of network. Control of AP s in customer premises and access network are controlled by this equipment. Also the user data coming from these AP s is passed via this Controller. In respect to physical medium, fixed networks use copper and fibre in the aggregation networks with the tendency of increasing fibre usage considering the ever-growing traffic. On the other hand, the mobile operators can choose microwave in addition to copper and fibre in their aggregation networks. The usage of copper in the aggregation networks is likely to decrease due to its limited capacity support and their inability to scale in a cost efficient manner. Looking forward, fibre is expected to take the place of copper based wire-line connections, and increase its overall share in the aggregation network infrastructure. Moreover, in the aggregation network of mobile operators, microwave is frequently deployed in developing markets and in emerging markets in which fibre is not available. In fact, at distances over several hundred meters, Microwave offers a much better cost-per-bit ratio than fibre. Microwave is also frequently used in developed markets as an alternative to costly line-leasing services offered by Telecom incumbents. Like FMC at the aggregation segment, as performed in Wi-Fi networking, the data plane at the mobile backhaul network can be handled as an usual service provider s traffic and therefore it is aggregated with Wi-Fi and fixed traffic to be transported and then switched to mobile core. Also regarding transport protocols, in order to achieve the same performance and quality of service, besides the widely used UDP and TCP, emphasis needs to put on the deployment of the reliable and message-oriented SCTP transport protocol at all network entities, as described at the end of the following section Core In the core network, the core elements for Fixed, Wi-Fi and Mobile are essentially different and do not share, e.g. common policy and changing functionalities. In the mobile core network the network entities and their functionalities are separated according to different aspects. In case of 2G/3G RAN, the two main domains are Circuit Switched core for voice and Packet Switched core for data. Incoming traffic is already separated at the aggregation level. In the CS core control functionalities are performed by the MSC and the GMSC; user voice traffic is routed through the MGW and the GMSC. In the PS core both control and user traffic are handled by the SGSN and the GGSN. Authentication, authorization, accounting and QoS are supported by the core network registers: HLR, VLR, EIR, AuC, etc. No final report - Waiting for acceptance from the European Commission Page 36 of 119

37 In case of 4G/LTE the Evolved Packet Core handles all user traffic as packet data, and an always on IP-connectivity is provided between the UE and the PDN. The network entities are separated by control and user functionalities. The control plane includes the MME with support from the core network registers: HSS, PCRF, CSCF, etc. User packets are routed through the SGW and the PGW, the latter being supported by the core registers. In contrast to the independent mobile core network, the fixed and Wi-Fi networks share a common core network architecture. User traffic is routed through switches and routers. Authentication, authorization, accounting and QoS functionalities are performed by dedicated AAA servers and BRAS. Control and network management is performed on the common network connections as in-band signalling (through the Radius and Diameter protocols). Current large scale Wi-Fi networks, whether they are of public or community types, also have a core platform. For distinction, this platform can be named as Wi-Fi Core. This type of core platform is not standardised in the way the mobile core is (e.g. 3GPP), and therefore current deployments are custom made when taken into account all the services included. The Wi-Fi Core provides a group of services where each can be hosted in specific hardware equipment and allows operating a large scale Wi-Fi network. The following elements can be considered to be present in a platform like this: AAA related systems (e.g. Authentication related servers, which can include a captive portal server for the web based authentication case). Network management systems (e.g. AP controller). Logging facilities. Security related systems (e.g. Firewalls and load balancers). Lawful interception systems (e.g. LNS server). Different authentication methods have been currently deployed, being the captive portal based user and password combination one the most relevant up to date. But even so many combinations are possible. In the area of network management as well several proprietary solutions have been developed by device vendors for Wi-Fi networks. Nevertheless and due to the lack of standardisation, they are not interoperable and are only valid for single brand deployments. For Mobile networks, integrated operational support systems are available providing functionalities for network and configuration management, allowing e.g., alarm surveillance of the nodes. In case of a convergent FMC network, emphasis needs to be put on the strict requirements for policy and charging functions being at the core of the mobile communication networks. Regarding protocol layer, in today s mobile core networks IP-based Sigtran control traffic is carried by the reliable, message-oriented SCTP transport protocol. Achieving the same performance would be difficult on the UDP/TCP transport protocols used in the current fixed packet data networks. Thus, in No final report - Waiting for acceptance from the European Commission Page 37 of 119

38 a convergent FMC network the deployment of the reliable SCTP transport protocol at all network entities seems of paramount importance Analysis according to functional groups The comparative analysis is here performed according to the eleven functional groups defined by COMBO. It thus addresses the differences between fixed, mobile, Wi-Fi networks for the implementation of functions in each functional group Forwarding As shown in Figure 14, structural convergence is already realized in today s aggregation network and IP backbone. This basically means that the same mechanisms are used for forwarding the traffic of the three network types (fixed, mobile and Wi-Fi) in these network sections. The IP backbone works with L3 forwarding on the basis of IP routing and MPLS whereas the aggregation network typically runs with L2 forwarding on the basis of Ethernet switching, MAC learning and VLAN handling, in some cases complemented by MPLS. However, the mobile network is an overlay network where the user IP data are tunnelled through the aggregation network and the IP backbone via tunnelling protocols (e.g. GTP) on the OSI layer-4. In the fixed access and Wi-Fi network the user IP data are tunnelled on layer-2 via point-to-point protocol (PPP) over Ethernet through the aggregation network or transmitted directly over Ethernet (IP over Ethernet). In addition, the Layer-2-Tunneling-Protocol (L2TP) is used for wholesale applications in order to tunnel the PPPoE session through the IP backbone to the location of a service provider. Multicast forwarding mechanisms are widely implemented in today s fixed access networks on the basis of the Internet Group Management Protocol (IGMP). In contrast, multicast forwarding is currently not used in the mobile network Automatic Configuration and Management Automatic configuration and management is a functional group that is addressed in various ways in the networks. Here are examples of technologies for respectively the mobile and the fixed network. For the mobile network: ANDSF: a help for automatic best network selection on UE side ANDSF (Access Network Discovery Support Functions) is a 3GPP technology used for helping the end user device to automatically select the best available access (Wi- Fi, 3G, LTE), based on real time data from the network (location of access points, their load in real time etc.). It is a form of convergence as it applies control to both Wi- Fi and mobile accesses. In the case of the fixed network, two mechanisms provide automatic configuration and management for CPE and network termination: No final report - Waiting for acceptance from the European Commission Page 38 of 119

39 TR69: autoconfiguration of the CPE device: The BBF TR69 specifies the CPE WAN Management Protocol (CWMP), intended for communication between a CPE and Auto-Configuration Server (ACS). The CPE WAN Management Protocol defines a mechanism that encompasses secure auto-configuration of a CPE, and also incorporates other CPE management functions into a common framework. The ACS is a server that resides in the network and manages devices in the subscriber premises. CWMP targets automatic configuration for residential gateways, Set Top Boxes, Wi-Fi plugs etc ITU-T specifies the ONT management control interface (OMCI) [11]. This OMCI specification addresses the ONT configuration management, fault management and performance management for G-PON system operation and for several services, including G-PON encapsulation methods (GEM) adaptation layers, circuit emulation service, Ethernet services, voice services Both protocols can coexist in fixed access network. TR-142 [12] clarifies the role of each protocol for residential gateway with integrated G-PON network termination. These are only two examples and many other technologies exist for each access type. The point is that these are very different ways to do autoconfiguration and management in the network and for different purposes (here access selection and CPE configuration), so the gap for having convergent techniques is really important today Resilience Resilience related functionalities (see section 2.2.3) are included in current fixed, mobile, and Wi-Fi networks. Resilience functionalities are needed in network equipment shared by many customers, so currently, network elements inside the aggregation and core networks require resilience functionalities (e.g.: redundancy, load balancing, dual Power Supplies ). This requirement is common to fixed, mobile and Wi-Fi networks. In fixed network, many network elements such as Ethernet / MPLS switches, BNG, LER, and IP backbone routers contain resilience functionalities, for example hot standby Switching fabrics. The same situation happens with the mobile network, where the Ethernet / MPLS switches, CSG, and mobile core elements typically include resilience functionalities at some degree. Wi-Fi networks also need resilience functionalities, and these functionalities are more important as the network element is more centralised, like the AP controller, the Wi-Fi access gateway or the Wi-Fi portal Security The first set of functions in this group includes the protection of communications either between the user and the Access Server or between the operator s network nodes. These heavily depend on the network type. Authentication: In wired network, the network operator controls the access points and can link user identity to an access point and use that information in authenticating No final report - Waiting for acceptance from the European Commission Page 39 of 119

40 users. This implies that procedures to identify the user can rely on this in-build secure access link. In wireless networks, the presence of a communication is not easily linked to the user who can be nomadic or mobile, therefore requiring cryptographic techniques for authentication relying on passwords, X.509 certificates, or keys stored in tamperproof devices such as SIM Cards. In mobile networks, Authentication and Key Agreement (AKA) allows the network to verify the identity of the mobile user and also to derive a session key for further encryption. Encryption: In wired networks, eavesdropping of data flows within the network is usually not possible, except by the network operator. Traffic encryption is not mandatory, although IPSec can be used. In insecure domains, such as wireless networks, eavesdropping and data forging or tampering is possible, making encryption necessary, generally by using IPSec, TLS, or the UMTS and LTE encryption algorithms. In the case of mobile networks, the AKA procedure is also used to perform session key agreement. The second set of functions aims at preventing and monitoring unauthorized access and misuse of the network resources. There are no significant differences between different network types, beyond the methods described above OAM & Management Operations, administration, maintenance and management in the mobile networks are carried out by dedicated equipment as well as functionalities integrated within the network entities themselves. Faults at the network s physical, data link, network and transport layers are continuously monitored and reported. Faults need to be corrected within the time frame as directed by the QoS agreements and business interests. This is also the case in fixed/wi-fi networks, if such QoS agreement (SLA) are in effect. In case of best effort services the fault management requirement are less strict. For the successful deployment of a convergent MFC network, fault management needs to support the strictest requirements set by the constituting fixed, Wi-Fi and mobile networks. For configuration, accounting, performance and security management, the same principles apply to all existing fixed, Wi-Fi and mobile networks; these considerations need to be taken into account in the FMC networks as well Synchronization Synchronization functions can be instantiated in a variety of technology dependent synchronization modules, according to: The type of synchronization to be provided, i.e. frequency/phase/time. The methods used for synchronization extraction and for synchronization distribution: synchronous physical layer (only for frequency synch), packet No final report - Waiting for acceptance from the European Commission Page 40 of 119

41 based (frequency / phase / time), GPS (frequency / phase / time). It should be noted that a synchronous physical layer method can also be used in conjunction with a packet based method. Synchronization modules based on physical layer methods differ by the physical layer type: E1, T1, SDH, xdsl, GPON, XG-PON, and Ethernet. Beginning from 2G, mobile radio access networks have been having strict synchronization requirements which may vary according to the multiplexing techniques used by mobile base-stations and also user application needs. Since demands on synchronization are at the edge of the access network, to satisfy these requirements time/frequency has to be carried over backhaul network to the basestations, if expensive GPS synchronization technique (GNSS) is not chosen.tdm based T1/E1 or SONET/SDH backhaul connections which are used in 2G-3G networks; carrying synchronization is not an issue because these networks are synchronized by their nature. Only synchronization of the backhaul network matters. Packet based backhaul networks do not require synchronization issues however time/frequency or phase has to be distributed over them if their associated access nodes are required synchronization over them. For example, access technologies like GSM, WCDMA, LTE-FDD and femto-cells only require frequency synchronization. This is optimally provided by the packet based backhauling networks with Synchronous Ethernet (SyncE) which provides SDH equivalent synchronization by deriving the timing from the physical links. The key drawback of SyncE is the need to implement it all across the network. However other technologies such as LTE-TDD require, in addition to accurate frequency synchronization, a high level of phase synchronization. As SyncE is not capable of providing phase sync, operators can deploy the IEEE1588v2 (PTP) technology or GPS. In DSL based fixed access network, the DSLAM and CPE needs only frequency distribution. In such deployments when ATM network exists behind DSLAM, frequency is obtained from the aggregation network. For the deployments like packet based aggregation networks behind DSLAM, packet-based protocols like SyncE or PTP can be used in DSLAM. For Wi-Fi networks, since at the access part APs do not need synchronization and fixed-based network composes backhaul of Wİ-Fi networks, the synchronization requirement of this network is related to underlying fixed network. Although synchronization is used to align device within the network, end user applications may exist such as mobile backhaul, financial services and SLA monitoring in order to measure round trip time accurately for example. Consequently, next generation mobile network backhaul technology is moving from circuit-switched to all packet-switched. Supporting all radio access technologies means maintaining both type of circuits that leads to increases in costs. Additionally regarding FMC that enables multiple technologies in the same network, having single type network, that is packet-switched based, becomes more logical in terms of cost and management. But this type of packet-based network has to emulate circuitswitch to provide backward compatibility to radio access networking. Hence while evaluating to packet networks in aggregation/backhaul segment that do not need No final report - Waiting for acceptance from the European Commission Page 41 of 119

42 synchronization, to distribute system clock to access edge nodes packet-based protocols like SyncE or PTP can be used Policy and Charging Fixed Mobile Convergence includes the chance for a subscriber to access services over fixed and wireless networks. Since 2012 Broadband Forum (BBF) addressed the FMC convergence by defining jointly with 3GPP business requirements and an interworking reference architecture, able to support extensions to BBF specifications and facilitating the interworking with the 3GPP Evolved Packet Core network (see TR-203 [13]). At the same time, Broadband forum has also filled a gap by augmenting the broadband multi-service architectures, with the definition of an integrated approach to policy management and control, i.e. Broadband Policy Control Framework, not clearly specified so far (see TR-134 [14]). The FMC Interworking architecture addressed by [13], aims to support service providers offering both fixed network access and 3GPP wireless access network, as well as separate access providers offering either fixed or 3GPP wireless access types and offering the other access type in conjunction with another provider. The policy decisions are taken in the 3GPP wireless domain and pertinent information needs to be transported via appropriate interfaces to the logical functions in the BBF domain. The following assumptions are made: Policy information will be derived from an Application Function (AF) in the 3GPP wireless domain. The pertinent information from the 3GPP AF will be submitted by the 3GPP PCRF to the BPCF via the S9a interface. All parameters/attributes transported by the 3GPP PCRF to the BPCF via S9a are as defined by 3GPP protocols. For authorization and authentication of a 3GPP user on a BBF access network, the 3GPP AAA infrastructure and BBF AAA infrastructure will exchange parameters using the STa or SWa interface as appropriate. Informational elements exchanged between PCRF and BPCF are also defined by TR-203 (in both directions) in terms of QoS, bandwidth profile, priority, IP addressing, UE identification. FMC is, thus, well addressed by the SDO s interested to the definition of fixed and wireless (mobile) networks for the aspects related to policy and charging. This activity is underway both in terms of consolidation of requirements (e.g. nodal requirements) and the definition of an IM suitable for the control of interworking policy and charging functions. The consequent step is the implementation into the fixed-mobile networks of the findings coming from the above standardization activity Subscriber Data and Session Management Fixed Mobile Convergence in subscriber data management is a way to focus the services and their quality to the customer benefits, independently from the access. If historically different subscriber management technologies emerged in each access type, OTT s show for years that it is worth for a service to be only linked to the No final report - Waiting for acceptance from the European Commission Page 42 of 119

43 customer rather than to the access type, except for the control of the access itself. In all access types, the access control use similar techniques from an access type to another. Here are the main similarities and differences between fixed, mobile and Wi-Fi networks in a subscriber data management point of view. AAA is omnipresent: An important network node in subscriber management is the AAA. The equipment supporting AAA are used in fixed broadband networks for internet access and VoIP, in Wi-Fi hotspots and in mobile networks. In mobile and IMS networks in particular, a HSS (or HLR before LTE) is used but it is based on Diameter protocol which is an evolution of the AAA Radius protocol, so HSS is just a particular AAA. So the common point is that all subscriber management databases are based on a AAA somewhere. Some AAAs deals only with the authentication whereas some include other functions such as accounting, and each one use a particular authentication method. But this shows that if the convergence is done in the network infrastructure, having a unique AAA/HSS subscriber database for managing the access control of all network types is relevant Mobility Mobility with respect to access networks implies functions which are required to provide a seamless service for moving customers. Traditionally, mobility is an issue and functionality only considered in mobile networks. Mobile networks such as GSM, UMTS and LTE therefore support a set of functions including measurement of the signal strength at the end-user device, reporting to the network and decision making. As a result, a handover from one base station to the next without a service interruption is state of the art in mobile networks. With growing availability of Wi-Fi access evolving towards the support for moving Wi- Fi users between multiple access points and access routers, Wi-Fi mobility has to be considered. Especially in view of interoperation between fixed wireless and mobile networks (e.g. off-load) mobility support within Wi-Fi and mobile access will be required. New mobility features within fixed networks may become important e.g. in the framework of cloud computing and virtualization: Here services such as data centre processing and content hosting may be operated at different physical entities which change during time of day due to capacity reasons or operational costs (e.g. energy cost). During an on-going session the location of an end point may change (which may or may not include address change depending on the implemented solution) and corresponding mobility functions have to take care of seamless service continuation. Mobility in current and future cellular (3GPP) networks is GTP and MIP based (latter one especially for including Wi-Fi as access) with new both user behaviour based and application and service specific requirements towards session and flow based rerouting and optimization. For fixed/mobile convergent networks, mobility may be applied as an inbuilt functionality with flag to switch on/off, e.g., for stationary mobile No final report - Waiting for acceptance from the European Commission Page 43 of 119

44 nodes. Also, the mobility support should be adapted to the provided services and the user s situation, instead of the used device and access technology. Therefore, supporting mobility can be envisaged in a more distributed and dynamic fashion. One important step will be to provide flexible mobility anchoring functions that can be activated on demand and distributed as close as possible to the user s terminal, e.g. at the access router level. The approaches developed in the framework of the DMM (Distributed Mobility Management) working group in the IETF can be considered and further evaluated. More details are provided on the handover management section (see section 5.4) Legal Interception and Data Retention Legal Interception (LI): The principle of the legal interception is the same for fixed and mobile networks. Via a mediation platform LI request from the Law Enforcement Agency (LEA) and LI responses from the operator are transferred. In the fixed Network the LI capability is realized mainly at BRAS. In the mobile networks different core network element are involved depending on the mobile generation and CS and PS domain to provide the LI response. In LTE e.g. the SGW and PGW are involved. The main differences are given by the network elements which collect the LI data and the locations where it is done. In the mobile network the elements are very central while the BRAS is more distributed in the network. Data Retention: The principle of Data retention is the same for fixed and mobile networks. The main differences lay in the network elements which collect the data and the locations where it is done. A main element in the fixed network which collects data is the BRAS. In the mobile network core network elements are involved collect the data. These data are temporarily stored (for max. 7 days) both in fixed and in mobile networks to request again if necessary. Emergency Call: In the fixed network, the identification of location of the emergency call for the All-IP VoIP customers is based on the IP address of the subscriber, while for POTS (MSAN) technology the connection is identified by dedicated identifiers (phone number and LineID). With these identifiers, the necessary location information can be detected. Depending on the technology of the inquiry terminal all these information (address, phone number, etc.) will be processed and then transferred to the jurisdiction of local inquiry terminal. In the mobile network the telephone service network provider determines the location of the emergency call ends - including the automatic emergency calls from vehicles During the execution of the emergency call both the phone number and the location data of cell tower are transmitted. The transmitted information contain a unique reference coordinate information to the antenna site based on circular ring segment related to geographical latitude and longitude coordinates in degrees, minutes and seconds of arc (acc. to ETSI TS ). The location information is stored in the parameter "User to User Information" (UUI) and continued in accordance with the call connection to the emergency call center. No final report - Waiting for acceptance from the European Commission Page 44 of 119

45 Traffic Analysis The tools available for traffic monitoring can be considered as part of the Operation Administration and Maintenance (OAM) framework. More advanced tools are implemented by inserting extra packets in the data flows and analysing the contents of these measurement packets. Others rely on inserting test traffic in order to probe network performance. A last category of tools relies on sniffing actual traffic (e.g. copying either packet headers or complete packets) at specific interfaces. There are no major differences between the methods used for the 3 types of networks. 3.2 Identification of gaps towards FMC networks This section identifies some gaps towards convergence, based on the use cases defined by WP2 in D2.1 [1]. The analysis is thus performed according to the four groups of use cases developed by WP Unified wireless access networks Use-cases UC01 FMC access for mobile devices describes a mobile broadband network in which mobile data traffic is offloaded to a Wi-Fi network when the user is located in an area with Wi-Fi coverage. UC02 Enhanced FMC access for mobile devices enhances UC01 by enabling mobile devices to simultaneously use both Wi- Fi and mobile accesses. Mobile devices can seamlessly move all or part of their traffic from one access to another, with the possible network assistance for selecting and using the most suitable access(es) according to their needs. Both UC01 and UC02 deal with the ability of a mobile device to use multiple wireless technologies, and therefore multiple access networks, according to different criteria (operator preferences, cost, network status, technical efficiency ). Wi-Fi technology is expanding at rapid pace nowadays and appears to be a valid candidate as a complementary technology in the near future to achieve cellular wireless radio access off-load, even though early Wi-Fi standards lacked the elementary features required for mobile data service. Users of both RANs (Wi-Fi and cellular) are faced today to relatively poor information about radio environment displayed by their devices (except for received signal strength indication and some network identifiers, such as Wi-Fi SSID or 3GPP Cell ID). In addition, switching from cellular to Wi-Fi often requires using a different identity and specific credentials before getting a full IP access. From UC01 and UC02, we can see that the user QoE is definitely improved in a way that network attachment is made seamlessly. To achieve such a goal, an additional function is required on the user device: a Connection Manager which is in charge of handling and enforcing static policies to select the most suitable access according to the operator policy and user preferences. Currently, capabilities of existing Connection Managers vary depending on the underlying Operating System capabilities. Third-party Connection Managers No final report - Waiting for acceptance from the European Commission Page 45 of 119

46 can also be added as add-on software. However, this field is not yet standardized. An additional protocol may also be required on the user devices and Wi-Fi APs to provide the device with some local and dynamic information regarding the network conditions. This is the goal of IEEE u protocol as defined by the Wi-Fi Alliance HotSpot 2.0 specification and PassPoint certification, which enables the AP to advertise such information (please see [15][16]). Moreover, once the selection process is completed, seamless access to a Wi-Fi access point is performed by WPA2-entreprise security for mobile SIM-enabled devices such as smartphones and also for non-sim devices such as low-end tablets or laptops (in the latter case, another EAP method, such as EAP-TLS will have to be supported, please see COMBO D2.2 [2], section 2.3 for details). WPA2-entreprise secured access has to be implemented in Wi-Fi APs, which is the case of most of commercial products at the time of writing. In order to enhance the access selection process, a dedicated entity on the network side will push dynamic policies to the user device connection manager to also consider its current location. Such a functional entity, performing Access Network Discovery and Selection Function (ANDSF) is currently under specification by 3GPP standardization body (please see [17]). The Connection Manager module will then have to process all the available information to determine the best available network according to both the static and ANDSF-provided dynamic policies. UC02 brings some additional constraints compared to UC01: seamless handover and double attachment. Seamless handover between Wi-Fi APs and between Wi-Fi and cellular mobile access requires a Mobility Manager entity in the network. Many network protocols mainly specified by IETF and 3GPP are already designed to cope with IP mobility. For limiting the required enhancement of the user device, a networkbased mobility control is suitable and may be achieved by a protocol such as IETF PMIP (please see [18], [19], [20], [21], [22]). Additionally dual interface mobile devices under the coverage of both cellular and Wi-Fi networks could be connected to both networks, using them at the same time according to the static and dynamic policies previously commented. While UC01 and UC02 are more network-function-oriented, UC03 Converged CDN for unified service delivery presents an enhanced content delivery architecture. UC03 shows the functional infrastructure and current / potential solutions of content delivery in fixed networks and mobile networks independently. The target is to find a way to provide unified content delivery solutions. CDN networks have been successfully deployed in current fixed networks. The recent Cisco VNI report [23] has shown that about one third of Internet traffic is carried by CDNs in 2012, and by 2017 the traffic crossing CDNs will grow to 51%. With the popularity of smart phones and emerging mobile applications such as healthcare and multimedia, mobile Internet is dramatically expanding. According to recent studies [23], Internet traffic from wireless and mobile devices will exceed traffic from wired devices by 2016 and nearly half of Internet traffic will originate with non- PC devices by then. To meet the growing data demands of mobile users, commercial LTE networks have been deployed to significantly increase the bandwidth and reduce the latency in the backhaul network. Though LTE improves network No final report - Waiting for acceptance from the European Commission Page 46 of 119

47 performances between end-users and mobile network, the service latency still highly depends on the distance of the data centre to the mobile network operator s point connecting to the Internet. This service latency is a key to satisfy the QoS of current applications, especially for latency-sensitive applications like audio and video services. Moreover, LTE network backhaul bandwidth will be heavily consumed by the duplicated content when this content (high-popular video) is requested simultaneously and frequently. Mobile network caching could be a cost-effective solution to improve the service latency and reduce the mobile backhaul traffic by replicating popular and frequently demanded content in the IP-based 3G/4G network elements closer to mobile users. Mobile network caching is getting a lot of attraction from researchers and companies. New infrastructures, cache framework and cache algorithms in mobile network have been studied [24], [25], [26]. Caching solutions for 3G or LTE networks have been developed by companies like Allot Communications, PeerApp, Altobridge, etc. Considering the mature CDN markets in fixed networks and emerging interests in mobile CDN networks, it would be worthy to start now looking at the possible methods for a converged CDN solution in FMC. A converged CDN could allow network operators to enable a better QoS, reduce internal costs and offer new content services Access resource sharing Convergence of fixed and mobile networks can provide substantial savings for the access network, especially if we consider sharing the access infrastructure. There are three of the COMBO use cases addressing access resource sharing: UC04 Reuse of infrastructure for indoor small cell deployment; UC05 Effective backhaul deployment for outdoor small cells; UC06 Common fixed and mobile access termination in hybrid connectivity for FMC customer service. UC04 and UC05 are related to structural convergence. For UC04 the main driver towards a Reuse infrastructure for indoor small cell deployment, i.e. for recycling existing residential and business indoor copper and fibre setup when deploying indoor small cells is to save cost and deployment time. This makes it a cost efficient and easier way to deploy indoor small cells. In addition, existing fixed copper and fibre infrastructure will be used more efficiently resulting in potentially lower costs both for the operator and for end-users. The goal of UC05 is to quickly and easily deploy backhaul/fronthaul connection for outdoor small cells. The suitable point-tomultipoint wireless NLOS (non-line of Sight) topology is typically hub-and-spoke with the possibility to extend the spokes with additional hops. In order to reuse existing infrastructure, it is beneficial to co-locate the hub with a macro cell site and/or a fixed infrastructure distribution node. It gives an operator more flexibility by providing yet another option for backhaul. Those two structural convergence use cases exhibit gaps in terms of how to provide required carrier-grade performance on the last-but one mile when utilizing potentially No final report - Waiting for acceptance from the European Commission Page 47 of 119

48 less reliable network infrastructure For UC04, in a typical office building the physical network is often privately owned and therefore out of control of operator. UC05 also covers residential buildings and the technology is strongly propagation dependent, since it uses high radio frequency for high bandwidth. Several proposals to achieve quality of service on unreliable (fixed or wireless) transmission links have been discussed mainly in the area of radio backhauling. Investigations providing general principles, key system parameters and engineering guidelines for deploying microwave backhaul have resulted in trials demonstrating that high-frequency above-20-ghz systems can outperform those using sub-6-ghz bands - even in case of missing direct line of sight (LOS) [27]. A solution enabling backhaul over any media (copper, fibre, or microwave) supporting efficient, scalable and reliable multichannel implementation combined with packet microwave ITU-T G.8032v2 ring network topologies increases link capacity and availability at affordable investment [28]. On the other hand ongoing and completed EU projects address how to provide a reliable transport network backhaul for cellular and wireless access nodes using different radio access technologies such as Wi-Fi over Long Distance (WiLD) for the backhaul [29], [30]. Here properly designed routing protocols and gateways are in charge of linking transport and access network segments to achieve the required transmission quality. The target of UC06 is to provide optimum bandwidth dynamically and resource efficiently to demanding customer via available fixed, mobile, and wireless technologies. Bundling is done on several different transmission paths (e.g. fixed DSL, WLAN hotspot, cellular radio). The service performance in this case depends on degree of structural network convergence realized between customer residential gateway and a hybrid connection proxy. To be able to allocate resources as flexible as possible on a per packet (MPTCP Multipath TCP) or per flow basis a similar delay on all transmission paths is often required, especially when handling TCP based traffic. From a protocol point of view there are several options such as Multipath-TCP [31], [32] and Multipath-STCP as well as proprietary solutions. [33] Introduces a proxy-based inverse multiplexer (PRISM) enabling TCP to efficiently utilize available wireless network connections while reducing performance degradation. In [34] a multilink network layer proxy is designed to transparently stripe traffic destined for multi-homed clients. Several extensions to MPTCP are under discussion. A proposal to cope with delay variations by using a congestion window adaption algorithm for the MPTCP source (CWA-MPTCP) is described in [35]. More proposals can be found at the corresponding IETF WG site [36] Aggregation resource sharing COMBO activities have already highlighted the role of the aggregation network in fixed/mobile convergence. This role is quite naturally performed by this portion of the end-to-end network due its traditional transport function, agnostic by definition with respect to the different services carried and accommodating, at the same time, with both short and medium distances No final report - Waiting for acceptance from the European Commission Page 48 of 119

49 The evolution of services towards packets technologies has enhanced the options of implementation (natively TDM and optical), allowing for a variety of technological flavours able to satisfy, when carefully combined, the transport of any type of client services. In COMBO D2.1 [1] different network use cases have been identified with the aim to, basically, stress the need for a further integration between fixed and mobile backhauling/fronthauling networks, which allows optimisation of infrastructure and functions, taking into account current and future service applications. In those use cases, the following macro targets have been addressed: sharing of the network infrastructure, extending either the current aggregation domain towards the access (fixed / mobile) or enhancing the access domain extent towards the main central office: UC09 Convergent access and aggregation technology supporting fixed and mobile broadband services. sharing of network infrastructure enabling for dynamic allocation of connectivity resources among mobile, residential and business services: UC07 Support for large traffic variations between residential and business areas. definition of an universal gateway function, at the core central office (i.e. the edge of the aggregation network), integrating both fixed and mobile controller functions and allowing, when needed, for IP off-loading from mobile gateways: UC08 Universal Access Gateway (UAG) for fixed and mobile aggregation network. While the first two targets identify mainly the structural convergence to be achieved on the transport infrastructure, the third one shows also meaningful aspects related to functional convergence at the end of the aggregation network (i.e. at core central office). The fixed/mobile convergence should take into account, not only future green field deployments, but also the evolution, possibly seamless, of the current networks into the future convergent infrastructure. From this perspective, addressing two alternative options for the sharing of the infrastructure is definitely appropriate: the extension of current aggregation domain towards the access portion appears suitable for traditional transport operators, already deploying aggregation networks as rings or small meshes; the extension of current access domain towards the main central office is preferred in those cases where the PON based infrastructure have been deployed and the chance to extend the connectivity to the main CO is easily achievable. In both cases, the optical technologies play a basic role in matching the transport needs of different service types, specifically when delay or jitter is sensitive (e.g. as CPRI/OBSAI signals). The WDM flavours already available (CWDM, DWDM and WDM-PON) allow, consequently, for matching the expected convergence, even if No final report - Waiting for acceptance from the European Commission Page 49 of 119

50 commercial aspects need to be improved. The packet and/or TDM capabilities (supported in aggregation and PON-based deployments with different means) complement the needed building blocks at the network nodes, performing the adaptation of client services into the server technology and optimizing the bandwidth usage. In addition, the future network is, expected to follow the support of new services by implementing dedicated adaptation blocks for the optimization of transport (e.g. CPRI/OBSAI over ODU). In the addressed convergent network, the capability to dynamically use connectivity resources, depending on the traffic demand over time, allows to reduce the power consumption and to minimize the CAPEX of the network by, possibly, sharing the same assets among different services in different periods of the day. Example of this approach is the CRAN model where BBUs are centralized at central office. CRAN allows optimization in the amount of baseband units, moving from a one to one relationship between BBU and RRU to an architecture where there are less BBU units than RRUs: BBUs are shared with an identified pool of remote radio units (RRUs). As far as the network infrastructure is concerned, the technological means currently existing would support since now the implementation of this sharing, provided to adopt the suitable means for supporting connectivity changes (e.g. switching functions and/or dynamic activation of optical resources). Finally, the definition of a universal gateway element able to control and monitor the shared network infrastructure completes the picture proposed by COMBO D2.1 network use cases for the aggregation network sharing. This block expected to be located closer to the final user (e.g. at the edge between aggregation and IP backbone) should realize the actual functional convergence by integrating the controller functionality related to both fixed and mobile, as well as Wi-Fi networks. With this approach, suitable in those cases where traffic does not need specific treatments (e.g. per user DPI), the efficiency of network control, maintenance and logistics would be improved with a resulting cost optimization both in terms of needed network resources (CAPEX) and in terms of operational costs (OPEX). This unified and integrated gateway would also allow for the dynamic allocation of connectivity resources across the network, complementing the target features above addressed (see bullet 2) and representing one of the basic aspects associated to the NG-POP concept. The gaps towards convergence, as far as the aggregation resource sharing is concerned, are as follows: Optical assets cost reduction; Implementation of means for ensuring dynamic connectivity (expected to be at L1, L2) over a single network for all kinds of services; Implementation of adaptation functions, depending on the services to be carried (e.g. CPRI/OBSAI over ODU); Adoption of new technologies finding the optimum balance between access and aggregation consolidation level; No final report - Waiting for acceptance from the European Commission Page 50 of 119

51 Integration of universal GW functions (fixed, mobile, Wi-Fi controllers, BBUs hotel, possible proxy functions for DCN etc.) in a unique network element, including appropriate means/procedures for security (e.g. appropriate handling of IPsec). Different integration profiles can be considered, in terms of relationship SW applications / HW resources Operator cooperation The multi-operator aspects in a possible FMC network are addressed in the UC10 dedicated to network sharing, as shown below. License supplier Site supplier Backhaul supplier Equipment supplier Services supplier Wholesale Network Company Provides shared radio capacity Owns active radio network Owns (or leases) sites/backhaul Consolidate sites Uses spectrum licenses of operators /site/month + /customer/month Asset transfer /site Operator X Operator Y Operator Z Figure 19: UC10 Illustration of network sharing using a third-party network company UC10 promotes the cooperation between operators in order to push the convergence of fixed/mobile networks. In the optical fixed network, the regulators already enforce network cooperation by sharing the optical cables from customer premise up to a mutualization point (located in the building or in a street cabinet). New fronthaul mobile offers are currently under development. Fixed operators will provide leased lines to mobile operators. These lines could be based on dark fibre, or on a dedicated wavelength, or on a complete managed solution. For legacy backhaul wholesale offer, the challenge will be to make these offers more flexible, to improve the cooperation and the convergence. The idea is to introduce some network virtualization: the operator could delegate some network management (link, equipment) to the operator which buys such wholesale offer. Network virtualization (or network slicing) introduces requirements on network equipment in terms of resource isolation and in terms of bandwidth isolation (need of hierarchical QoS for example). Coupled with SDN mechanisms, the third party operator could be able to apply its policies in a very flexible manner. Another area of exploration could relate to network virtualisation and its application towards RAN sharing between operators. No final report - Waiting for acceptance from the European Commission Page 51 of 119

52 4 IMPACT OF NEW FRAMEWORKS After the bottom up approach of section 3, based on an analysis of differences between network types and the identification of gaps towards convergence, this section focuses on the impact of new frameworks (Cloud Computing, Software Defined Networking, Network Function Virtualization, Autonomic Networking) through a top down approach of network convergence. It describes the potential benefits of such emerging industry frameworks for network convergence, and analyses in particular if those new frameworks can help in solving the gaps to convergence identified in section Cloud Computing Cloud computing is an industry term that has been adopted in the last four years or so which reflects the adoption of computing and storage facilities typically in large datacentres that are used to run applications which are remote from the user. The user typically uses a thin-client to access the remote data and access the application. Cloud computing has generated a number of underlying concepts which are classed as service models: Software as a Service SaaS; Applications such as , office applications, virtual desktops operating remotely; Platform as a Service PaaS; Databases, web servers, development tool suites running remotely; Infrastructure as a Service IaaS; Virtual Machines, servers, storage and networking functions running remotely. Important considerations in relation to Cloud computing include the latency with which end-users will be faced, the movement of data between data centres for resilience and disaster recovery and also the notion of on-demand services. Cloud services are said to be elastic, that is the compute or storage resource might be needed at any time by the user and the network infrastructure therefore needs to cope with changing demands in order to support cloud computing. Under the general topic of cloud computing, the IaaS model is being considered by Cloud RAN proponents as a way of engineering the Radio Access Network in a more scalable, and dynamic manner. Thus the notion of supporting the compute portion of a base station within a server facility as a pool of Virtual Machines, and then transmitting the radio signal via a fronthaul enabled network that is also engineered for more general access backhaul is an obvious way in which cloud computing might enable convergence of fixed and mobile networks. No final report - Waiting for acceptance from the European Commission Page 52 of 119

53 Another side effect of engineering the network for cloud services, is that from the need to support high bandwidth, low latency, and traffic agnostic dynamic services, one can envisage a network that is prepared for transporting many different applications. The concept of identifying flows of traffic rather than forwarding packets on an individual basis could in some ways be attributed to the origins of cloud computing. A scenario of cloud computing that is closer to the end-user equipment is also considered below. Mobile applications and mobile devices (such as smart-phones and tablet computers) have recently become significantly popular. However, the resource limitations on the mobile devices (energy, memory, storage, etc.) and mobile networks (bandwidth, latency, connectivity, etc.) have hindered this vision. To support sophisticated applications (for example, advanced games) on mobile devices, the cloud computing that centralizes resources (e.g. storage, computing and services) as a utility has been studied and deployed to provide an immense and scalable platform for service delivery. In the cloud computing, mobile users need access to servers in the cloud data-centres that are usually far from end-users to use cloud services and cloud resources. For example, the mobile apps offload their components requiring large computing resource to the servers in cloud. MAUI [37] enables fine-grained energyaware offloading of mobile codes to a cloud by partitioning the application codes at a run time based on the costs of network communication and CPU on the mobile device. CloneCloud [38] introduces offloading execution from the mobile device to cloud through running a mirror image of a smartphone on a virtual machine in the cloud. This can help reducing the energy consumption and supporting some powerful applications, however, mobile users usually suffer from the insufficient network bandwidth, network disconnection and signal attenuation due to the mobility feature. It causes long network latency when end users communicate with the cloud centre, which can reduce QoS significantly. Resource-rich computers or powerful mobile devices that are well-connected to the Internet have been recently investigated to be integrated into the cloud platform, in order to reduce network latency and improve QoS for nearby mobile users. Instead of offloading to the cloud which introduces long WAN delay, low bandwidth and high cost, mobile users can use these nearby computers that can provide low-latency, one-hop, high-bandwidth LAN wireless access for users. This allows the resources to be positioned close to end-users, which can significantly improve the QoS of mobile applications, especially for real-time interactive applications. Authors in [39] propose to use a trusted and resource-rich computer or a cluster of computers so-called a Cloudlet to share its resources with nearby mobile devices. The cloudlet infrastructure is based on Virtual Machine (VM) technology. A mobile device delivers a VM overlay to cloudlet infrastructure, and then the infrastructure creates a launchable VM for the user. Authors in [40] further propose a component-based offloading to allow discovering and managing available devices in a LAN to cooperate in the cloudlet. The basic operation in mobile cloud is the offloading of mobile computation and data backup from resource constrained mobile devices to cloud. The offloading No final report - Waiting for acceptance from the European Commission Page 53 of 119

54 communication has several impacts on the network. First, Offloading to cloud introduces more bandwidth consumption, especially for upload link. Second, the offloading involves traffic communication from local devices to remote servers by 3G / LTE networks, or to Wi-Fi network accessing to the Internet by fixed networks (xdsl, FTTH, etc.). 4.2 Software Defined Networking All communication nodes have to implement forwarding rules, i.e. methods for selecting a given output link for each incoming packet. Forwarding rules may depend on input link, on characteristics of the relevant traffic flow and on policy rules specific to the network, and relative e.g. to authorization, security, QoS, etc. Different technologies use different methods for deriving these forwarding rules. Typically, forwarding rules in IP networks are derived from routing protocols, forwarding rules in MPLS networks are derived from the definition of LSPs, and Ethernet forwarding rules are dynamically learned by observing received packets and linking their source address with the used input link. What is common in all these methods for deriving forwarding rules is that the actual derivation process is distributed, and that equipment individually maintain their own forwarding table, i.e. the set of all current forwarding rules. More generally, the creation and maintenance of a forwarding table is part of the control plane. Current communication equipment perform both forwarding (data plane) and control plane functions; interfaces between control and data planes are proprietary, and depend typically on the version of specific equipment. Moreover, the expression of policy rules is also proprietary, which makes deploying global rules over a network a significant configuration problem. Software Defined Networking (SDN) relies on data plane and control plane functions being realized in different equipment. Data plane functions, such as policy enforcement or forwarding, are performed in network devices that are externally controlled; Control plane functions, such as access control or path computation, are performed in so-called SDN controllers, which maintain a global view of the network. SDN controllers pilot forwarding and other data plan functions thanks to a protocol such as OpenFlow [41] over the so-called southbound interfaces (interfaces between the SDN controllers and the network devices); Open interfaces (APIs) are made available on those controllers to allow the application layer to program the SDN controllers over their northbound interfaces. A schematic representation of these principles is illustrated in Figure 20 where Infrastructure layer is the data plane and Control layer is the control plane. The arguments presented in favour of SDN are numerous: No final report - Waiting for acceptance from the European Commission Page 54 of 119

55 Network devices should be much simpler that today s equipment, as they would not have to implement control plane functions; they would thus be cheaper and more versatile; Control functions, per se, are simple; but distributing these functions is complex. For example, Dijkstra algorithm for computing shortest paths can be explained in 2 pages whereas the current IETF OSPF protocol specification [43] contains more than 200 pages; Having all network devices controlled by a single protocol (such as OpenFlow) facilitates the configuration of global policy rules. In today s networks, configuring such rules across all networking elements is complex and errorprone, as their syntax is proprietary; New service deployments should be facilitated as they only have to be translated in software based rules into the SDN controllers via APIs. In today s networks, such deployments are complex. Indeed, they are based on language machine based implementations, specific to each networking element on the one hand, and instantiation of many control plane protocols on the other hand; Highly dynamic environments such a data centres, banks of virtual machines and virtual machines migration require versatile and automated control mechanisms, which are hard to implement with current networking technologies. Figure 20 : SDN principles according to the Open Networking Foundation (ONF) [42]. It is worth noting that similar arguments have been presented in favour of IMS and TISPAN. However, SIP interfaces did not control networking functions but service level functions. Therefore, SDN addresses lower layers than IMS or TISPAN addressed. A better analogy can be made with operating systems [44] that present abstractions of resources (e.g. memory, storage, etc.) and deal with information (e.g. files and directories) instead of their actual representation in machine specific language. Using No final report - Waiting for acceptance from the European Commission Page 55 of 119

56 these abstractions, software programs can operate on many different hardware machines. In order to operate today s networks, it is necessary to deal with network and domain specific information, instead of their abstractions: for example, an Access Control List requires the knowledge of the IP address used by a given host. The control plane provided by the SDN architecture can thus be compared with a Network Operating System, or Network OS. It can be used to design many control functions through the northbound interface; it is a (logical) central entity, which has to maintain the mappings between the network abstractions (the hosts, the application servers, etc ) and their network bindings (IP and MAC addresses, port numbers, etc ). From the point of view of fixed-mobile convergence, SDN could play a role both for functional convergence and for structural convergence. For functional convergence, the specification of the function would not depend on whether it is applied to fixed or mobile customers. The service primitives would be translated into functional blocks that would later be mapped thanks to the Network OS on the data plane. It would be up to the Network OS to maintain the bindings between users and (temporary) addresses; For structural convergence, the massive use of networking elements supporting the same controlling protocol (e.g. OpenFlow) would facilitate the sharing of the networking architecture between fixed and mobile networks. 4.3 Network Function Virtualization Building on a well established trend in the IT industry, virtualization is rapidly entering the Telecom industry, Network Functions Virtualization (NFV) being its latest incarnation. Leveraging cloud computing technologies, NFV is a topic of much hype at the current time, however due to the number of companies exploring the subject there is a strong indication that it may become a rapidly adopted paradigm in telecom operator networks. Currently, most telecom equipment is sold in the form of integrated vertical systems (sometimes referred to as network appliances) with applications running on purpose-built middleware and hardware. The proposition of NFV is a paradigm shift, from the current situation to a cloud model where telecom functions are virtualized and run on virtual machines in pools of commodity servers. This approach already found its way in the IT industry. What makes NFV a challenge in this application is the telecom industry s demanding requirements for five-nines availability and for predictable real time performances. No final report - Waiting for acceptance from the European Commission Page 56 of 119

57 Figure 21: Network Function Virtualization overview The term NFV was coined by a group of Tier-1 telecommunications operators and introduced via the ETSI standards organisation in a White Paper published in October 2012 [45]. Network functions virtualisation is defined as the process of implementing network functions in software that can run on a range of industry standard server hardware, and that can be moved to, or instantiated in, various locations in the network as required, without the need to install new equipment. Such locations include but are not limited to - highly-centralized data centres and smaller regional data centres in network points of presence (PoPs). ETSI-NFV delivered its first specifications in October 2013, dealing with use cases [47], architectural framework [48], terminology for main concepts [49] and virtualisation requirements [50] Use cases and expected benefits NFV is expected to provide network operators with CAPEX and OPEX savings as observed via the use of Cloud technologies in the IT industry [46]. This includes: Lower equipment costs brought by the use of Commercial-Off-The-Shelf (COTS) hardware (including savings on the chassis themselves); Expected energy reduction brought about by consolidation of functions into fewer boxes, leading to optimization of power supply resources and cooling systems; No final report - Waiting for acceptance from the European Commission Page 57 of 119

58 Reduced time-to-market to deploy new or upgraded network services (no new equipment required, just software installation/upgrade); Greater flexibility to scale up and down resources assigned to applications based on actual usage. This is also known as elasticity; Lower operational costs since multiple network functions can be combined on a single platform and network functionality can be executed wherever it is most effective and efficient e.g. in terms of resource and cost spending; Uniform equipment management reduces operational complexity, no adaptation process needed in order to manage many different network elements. Use cases for NFV include functionality for subscriber management, enforcement of security and quality of service per session or connection, and content-related packet processing functions. These are realized within logical network entities such as CDN controllers and caches, functional entities of the IMS and EPC architectures, BNGs, firewalls, load balancers, DPI equipment, base stations, residential gateways and MSCs. There is theoretically no limit to the type of network functions that can be virtualized, although every use case brings specific challenges. Special attention must be paid by operators to the performance of virtualised user plane equipment units and their underlying costs. This also applies to network functions close to the physical layer and/or having to support very high-performance data transport (e.g. core network routers) and/or requiring high predictability of workload scheduling in order to avoid latency or jitter (e.g. Radio Network Controller) for which integrated boxes and embedded Application-Specific Integrated Circuits (ASICs) may be still relevant. NFV is also expected to fundamentally change the way network services can be chained. In this context, service chaining refers to creating a sequence of network functions, each of which providing a network service (e.g. firewall, DPI, NAT ), that packets matching certain conditions have to traverse. Service chains exist in conventional networks and are relatively static, pre-provisioned at layer 2 or layer 3. NFV is expected to provide more flexibility and dynamicity in the way services are chained, by decoupling service delivery from the underlying topology. From the point of view of fixed-mobile convergence, NFV will play a role both for functional convergence and for structural convergence. For structural convergence, the same processing/storage infrastructure could be used for fixed, mobile and Wi-Fi gateways which in terms of NFV means universal servers hosting basic functional modules to be combined for each service in a flexible way. Network element based on dedicated hardware could also host virtual network functions; For functional convergence, NFV could facilitate the creation of new functional blocks which are common to fixed and mobile networks. NFV could also help to virtualise the fixed and mobile service creation and chain it together for hybrid access. No final report - Waiting for acceptance from the European Commission Page 58 of 119

59 4.3.2 SDN and NFV NFV has become highly associated with Software-Defined Networking (SDN). Though the two technologies address different needs and have different objectives, i.e. SDN aims at making the network programmable while NFV aims at relocating network functions from dedicated servers to pools of commodity servers, there is a an intersection between the two that remains to be clearly identified. The aforementioned White Paper [45] says that NFV is highly complementary to SDN, but not dependent on it (or vice-versa). Network Functions Virtualisation can be implemented without a SDN being required, although the two concepts and solutions can be combined and potentially greater value accrued. NFV can indeed benefit from SDN technology. Separation of control and data forwarding planes, a key ingredient of SDN architectures, can improve performances and simplify compatibility with existing deployments, by enabling virtualization of control plane functions while leaving the forwarding plane implemented on dedicated servers. SDN may also prove useful to control the communication between virtualized network functions and implement virtual networking capabilities in the NFV infrastructure. Conversely, SDN controllers and switches can benefit from NFV technology as any other network function. 4.4 Autonomic Networking Autonomic systems were first described by IBM in The fundamental concept involves eliminating external systems from a system's control loops and closing of control loops within the autonomic system itself, with the goal of providing the autonomic system with self-management capabilities, including self-configuration, self-optimization, self-healing and self-protection. With these self - functions, the aim of the autonomic systems or autonomic networking is to reduce network complexity, increase automation and therefore, to reduce OPEX. Autonomic Networking aims at putting the intelligence of today's operations back into algorithms at the node level, to minimize dependency on human administrators and central management systems. The autonomic control loop is made up of Monitor, Analyze, Plan and Execute, all of which rely on a common knowledge repository. The Monitor component gathers data, filters, and collates it as required and then presents it to the Analyze component, which seeks to understand the data and determine if the managed element is acting as desired. The Plan component takes these data and determines if action should be taken to reconfigure the managed element. The Execute component translates the planned actions into a set of configuration commands that can be applied to the managed element. Autonomic networks seek to drive the Services and Resources provided by the network through the use and application of the appropriate business rules, governed by policy. If the network were in fact to work autonomically, then the human operator has necessarily changed from a hands-on worker to more of an advisor. The properties of autonomic networking can be listed as follows: No final report - Waiting for acceptance from the European Commission Page 59 of 119

60 Self-aware refers to perception and cognitive reaction to an event or more generically to a condition, relevant to the same node (or component, or system) with respect to its environment. This context-awareness is fundamental to realize all other functionality; Self-locating establish and update a reference system to identify neighbors and to locate the resources required for a coordination schema; Self-configuring ability to dynamically configure itself (node, component, system) with information pervaded environment that can adapt immediately to changes (including deployment of new components or changes in the information-pervaded environment itself); Self-healing analyze state, evaluate condition and commit corrective actions without disrupting any operation. With self-healing functions, the whole system becomes resilient as changes are made to reduce or help to eliminate the impact of failing parts of it; Self-optimizing functionality that refers to the ability to effectively minimize resource allocation without compromising the overall operation. What is expected with this functionality in the long term is that components may learn from experience and be able to proactively tune their behavior in the context of the overall objective; Self-protecting provide the right information to the right consumers (or users) at the right time through actions that grant access based on role and associated privileges. This includes the detection of hostile or intrusive behavior and the commitment of appropriate actions against it. Standards related to Radio Access Networks: ETSI ISG AFI With the increasing interest on Autonomic Networking/Self-Managing Networks, a new special working group is established within ETSI, which is an Industry Specification Group (ISG) called Autonomic network engineering for the selfmanaging Future Internet (AFI). The main aim of this group is to seek to establish a common understanding on what an autonomic behavior is and how an autonomic/self-managing network should be engineered. The Autonomic Network Engineering for the Self-Managing Future Internet ISG aims to serve as a focal point for the development of common Specifications and engineering frameworks that guarantee interoperability of nodes/devices for Self-managing Future Networks and will develop ETSI pre-standards and specifications for Autonomic Network Engineering for the Self-Managing Future Internet. 3GPP LTE Standard The standard defines self-configuration and self-optimizing process as follows: Selfconfiguration process is "the process where newly deployed nodes are configured by automatic installation procedures to get the necessary basic configuration for system No final report - Waiting for acceptance from the European Commission Page 60 of 119

61 operation. This process works in pre-operational state. Pre-operational state is understood as the state from when the enb is powered up and has backbone connectivity until the RF transmitter is switched on". Self optimizing process is defined as the process where UE (User Equipment) & enb (enhance Node B) measurements and performance measurements are used to auto-tune the network. This process works in operational state. "Operational state is understood as the state where the RF interface is additionally switched on". As an example, the standard describes the Automatic Neighbor Relation Function (ANRF) which allows a cell to automatically identify its neighboring cells. Different use cases of self-optimizing networks have been studied such as: Coverage and capacity optimization Energy savings Interference reduction Automated Configuration of Physical Cell Identity Mobility robustness optimization Mobility Load balancing optimization Load balancing can be performed via auto-tuning of handover parameters of mobiles in connected mode or of selection/reselection parameters of mobiles in idle mode. Interference reduction is performed via Inter-Cell Interference Coordination (ICIC) process and has a particular importance in OFDMA systems (i.e. LTE, WiMax). Work on the architecture and signaling protocols in 3GPP that allows the implementation of the SON functionalities is in progress. IEEE 1900 Standard The IEEE 1900 standard uses different concepts in autonomic networking for future RANs in the context of Cognitive Radio. Cognitive Radio is defined by the Federal Communications Commission (FCC) as a radio that can change its transmitter parameters based on interaction with its environment with two primary objectives in mind: highly reliable communications whenever and wherever needed; efficient utilization of the radio spectrum. Impact to FMC Autonomic functions are distributed among the network elements that support it. These autonomic functions become additional functions for the corresponding elements besides their main functionalities. Architectures of autonomic functions may vary depending on the deployment like fully centralized, fully decentralized or hybrid. These architectures are applied by sharing responsibility between a central controller and distributed clients working on network elements. Autonomic functionality on networking elements generally responsible for getting performance measurements, No final report - Waiting for acceptance from the European Commission Page 61 of 119

62 communicating with other autonomic elements, getting self-configuration from management server and as a result, from all these gathered data, either taking executive actions or sending the relevant data to a central autonomic decision server which later takes executive commands and sends to the element to perform. All these communications are done through control and management plane of the network which can be both standardized or vendor specific. Currently, the autonomic networking functionalities are defined for both access networks as well as aggregation and core networks. The standardization efforts regarding autonomic networking have been started for fixed and mobile networks separately and they are still in its infancy. The impact of these standardization efforts will shape the design and implementation of autonomic networking functionalities within FMC networks. Depending on the level of convergence and the use case scenarios, what kind of autonomic functionalities and how to implement them in FMC networks might open new study areas and need new standardization efforts specific to FMC. It is seen that self- capabilities coming with autonomic networking, especially self-configuring and self-optimising, will be very vital functionality for FMC networking in handling cost and energy saving targets. No final report - Waiting for acceptance from the European Commission Page 62 of 119

63 5 SOME TECHNOLOGICAL ENABLERS FOR CONVERGENCE Whereas section 4 dealt with architectural frameworks and emerging paradigms, this section is focused on some technological enablers which are believed to be keys of future FMC networks. 5.1 Unified optical access & aggregation network One basic outcome of the use case discussion in D2.1 [1] is definitely the demand for structural convergence of access and aggregation segments over a common infrastructure: a convergent approach where, a unique network is able to collect both fixed and mobile services leads to a potential CAPEX and OPEX reduction, with different flavours depending on the specific application scenario (services, involved players, geographical area, regulatory policy). LTE networks are currently being deployed in a wide scale while the generalization of optical access remains in most European countries a long term challenge, especially outside urban areas. However most COs are connected by optical fibres. This enables to optically connect the enodeb with the closest CO. This optical connection provides both: high throughput (multi-gbits/ speed) and low latency which are needed for LTE and LTE-A operations (e.g. IC-IC and CoMP). In the present aggregation/access architecture, the CO is located at the border of the access and aggregation network. The location of the CO was determined by the physical limitations of copper network. However, this border may evolve with the generalization of optical access, which provides the operators an opportunity to lower their OPEX by reducing the number of COs. This trend will be fuelled by the introduction of long-reach PON technologies. In this context, if the fixed and mobile network infrastructures are shared, the enodeb could be in principle connected to optical access equipment. That would work for pure backhauling (bandwidth <1Gbit/s). In case of fronthauling (i.e. CPRI / OBSAI based) the bandwidth requirements (>>1Gbit/s) and the very strict delay / jitter requirements make the usage of TDMA based solutions critical and lead to the adoption of WDM based solutions. The future network is, consequently, expected to effectively support current services packet based (IP/MPLS/Ethernet), legacy SONET/SDH applications (e.g. for the support of legacy 2G-mobile services, still to be considered in the future applications) and new services and related enabling technology, as CPRI fronthauling in BBU centralization. The convergence among all of these services may be realized through TDM/TDMA technologies in combination with Coarse/Dense WDM technologies to scale and exploiting efficiently the fibre infrastructure. The figure below shows an example for achievable convergence among fixed (Wi-Fi) and mobile access realized through a unified optical access infrastructure, based on TWDM PON in overlay with WDM- PON, along with a multi-layer aggregation network. No final report - Waiting for acceptance from the European Commission Page 63 of 119

64 Note The showed number of segments in both Access /Aggregation and network elements is purely indicative and depends on the actual deployment scenario. ACCESS AGGREGATION CORE ONT/CPE Access (fixed) Business ONT/CPE Residential Power splitter RN DSLAM Unified Optical Access through TWDM PON + WDM point-to-point TWDM PON RN WDM band filters Aggregation Edge OLT Aggregation & IP Edge Core network Small cells IP CSG Multi-layer AggregationNetwork (DWDM, OTN, MPLS-TP rings or small meshes) Macro Mobile Note Extent of access, aggregation and core segments are notshowed in scale CPRI CSG Example of small cells backhaul through TWDM PON Figure 22: Example of Unified Optical Access and Convergent Aggregation Network by TWDM PON and WDM technologies A specific network element (NE) access dependent performs in the upstream direction, the first grooming of user services. Possible network elements include: Power / Wavelength splitter (RN), according to the GPON/XGPON/NGPON2 or WDM-PON distribution network; DSLAM; IP Cell Site Gateway, aggregating 2G, 3G and LTE cell site traffic; CPRI Cell Site Gateway, performing the consolidation of CPRI signals coming from 2G, 3G and LTE RRUs. In the access segment, the star topology (PON based) is realized. Wavelength filter elements allows for the overlay between TWDM and WDM PON. At aggregation edge location, a TWDM OLT system handles the services backhauled (from RN, DSLAM, IP CSG) interoperating, at the same time, with an aggregation edge node. CPRI CSG is assumed to connect directly to the aggregation edge node. No final report - Waiting for acceptance from the European Commission Page 64 of 119

65 An alternative approach is the deployment of ring based network in the access segment, as depicted in Figure 23. ACCESS AGGREGATION CORE ONT/CPE Business ONT/CPE Residential Small cells Access (fixed) connectivities DSLAM Macro xpon splitter IP CSG OLT Unified Optical Access through CWDM/DWDM rings (OADM/OA not showed) Aggregation Edge Aggregation & IP Edge Multi-layer AggregationNetwork (DWDM, OTN, MPLS-TP rings or small meshes) Core network Mobile CPRI CSG Example of small cells backhaul to macro through C/DWDM rings (OADM not showed) Note Extent of access, aggregation and core segments are not showed in scale Figure 23: Alternative approach to Unified Optical Access and Convergent Aggregation Network In this case the Unified Optical Access may be realized through ring topology with specific OADM blocks integrated in or co-located with the access dependent NE s. Depending on number of NE s and traffic rate to be collected, either DWDM (from 40 up to more than 100 channels in C, L bands with 100 GHz or 50 GHz channel spacing) or CWDM technology (potentially up to 18 channels) may be used. Depending on the path penalty to be matched, OA s (Optical Amplifiers) may also be considered. The adoption of bi-directional transmission on a single fibre, so as, the deployment of colorless/low cost tunable interfaces, can enable a further CAPEX/OPEX reduction. In both PON based and ring based solutions, WDM technologies play a basic role in realizing a convergent aggregation network. The deployment of small cells for enhancing cell capacity is widely considered as one of the drivers in mobile network evolution. The amount of small cells into a macro area is expected to be within 10-12, according to NGMN Alliance. For this massive roll out a low cost solution for backhaul or fronthaul is required. As above considered, also for this network application, one option is the backhauling via TWDM PON or the No final report - Waiting for acceptance from the European Commission Page 65 of 119

66 backhauling via small rings, possibly single fibre with either ultra-low cost DWDM or CWDM as multiplexing technologies. 5.2 Heterogeneous radio access networks The popularity of smart devices (smartphones, tablets, etc) has been changing enduser behaviours and the provision of ubiquitous and seamless connectivity became one of the major challenges for telecom operators and vendors. Heterogeneous networks can be seen as a solution for this challenge. It refers to the mixture of various radio base stations, comprising nodes with different coverage or transmit power capabilities (macro, micro, pico cells), associated with different types of transport network infrastructure (copper, fibre, or wireless backhaul links). Currently deployments consider the integration of base stations with low transmit power (so called small cells) into an existing network with existing high power macro base station, improving connectivity e.g. in areas where the macro cells coverage is not sufficient enough. A heterogeneous network deployment example, including small cell deployment indoor/outdoor in residential and business areas, is depicted in Figure 24, where the backhaul logical links are indicated for both macro and small cells layers. Figure 24: Deployment example for Heterogeneous Networks. The introduction of heterogeneous radio access networks will push a better integration of mobile and fixed access networks because of a much larger number of antenna locations in the network. It is also an opportunity for FMC functionalities, such as 4G/Wi-Fi seamless mobility/connectivity. Heterogeneous radio access networks are thus a technology enabler for FMC realization. In this way, the following aspects are examples of how technical requirements related to heterogeneous networks can support FMC: Need for cost-efficient small cell deployment, which especially applies to the backhauling infrastructure and technology. This can be achieved by using the No final report - Waiting for acceptance from the European Commission Page 66 of 119

67 complete portfolio of fixed access technologies and existing infrastructure which are available in the vicinity of the small cell, such as FTTx and xdsl (potentially with bundling). Small cells might be optimal when coordinated with existing macro cells. This can efficiently be implemented with the help of BBU hostelling (see section 5.3) and a C-RAN architecture like network, where the BBUs of a given macro cell and of its related small cells are located in the same place, thus allowing an easy coordination. Small cells can support the same technology as the macro cell network (e.g. LTE) or a different wireless family (e.g. Wi-Fi small cells). In the latter case a tight interworking between the radio technologies is required (see section 5.4), considering e.g. coverage of residential areas. For further considerations on small cells deployments, please see [51][52]. Regarding to aspect 2 above, macro cell and small cell layers coordination can be implemented by means of C-RAN and BBU hostelling, or by the use of distributed procedures that require the communication between base stations through X2 interfaces. These distributed procedures can be used to implement macro cell small cell layers interference mitigation techniques like eicic (enhanced Inter Cell Interference Coordination), which require a tight synchronization of the base stations with a common timing distributed by the fixed access backhaul, or LTE-A techniques like CoMP (Coordinated Multi Point), which demand a precise synchronization and the simultaneous delivery of coordinated data flows to many different base stations. This direct inter base station communication is a major challenge for the fixed section of the network, as it can potentially require high data rate, low delay and synchronized solutions in a cost effective way. As a matter of fact, the one-way latency (not including cable propagation) of an ideal backhaul, as currently specified by 3GPP for Release 12, should be less than 2.5 µs [53]. Another relevant technology for convergent heterogeneous networks is SON, or Self Organizing Network. SON is expected to reduce the burden of mobile network configuration and optimization, and it relies on the exchange of information between base stations, which is conducted by interfaces like LTE s X2, or the exchange of information with a centralized SON server, and in both cases supported by wired and wireless links in a convergent network. Although the requirements, in terms of data rate or delay, on these links are not usually very demanding, it is certainly required the existence of an ubiquitous connectivity potential, in the sense that all the base stations in a given area should easily find a communication link with each other or with the SON server. 5.3 BBU hostelling and mobile fronthaul technologies In this section, the application of BBU (BaseBand Unit) hostelling is introduced together with an overview of mobile fronthaul technologies. But first, the motivation behind the centralized RAN architectures is explained is four different steps. No final report - Waiting for acceptance from the European Commission Page 67 of 119

68 In the first traditional generations of macro base stations (Figure 25), the radiofrequency transmitter and receiver electronics are located at the base of a tower, or in a building, and large diameter coaxial feeder cables are used to connect the electronics and the antennas. RRU: Remote Radio Unit BBU: BaseBand Unit D-RoF: Digital Radio over fibre (CPRI or OBSAI standard) Figure 25: Step 1: Macro base station, all base station hardware is located in a radio site cabinet, and the antennas are driven through coaxial cables. RRH: Remote Radio head CSG: Cell-Site Gateway Figure 26: Step 2: Distributed base station with traditional backhaul, remote radio heads at the antennas are connected to the radio site cabinet through an optical fibre. In a second step (Figure 26), distributed RRH appears. The RRHs contain the RF transmit & receive components (including power amplifier, duplexer, low noise amplifier etc) and they can be mounted directly on the antenna mast thus only short coaxial jumpers are used for the connection to the antennas. Due to their lower No final report - Waiting for acceptance from the European Commission Page 68 of 119

69 capital and operating expenditures, these RRH are currently being deployed not just for new technologies (e.g. LTE) but also in new and replacement infrastructure for older technologies (2G, 3G). The RRH can be linked to the Base Band Unit (BBU) by an optical single mode fibre using a standard interface with a digital radio signal [54][55][56][57] (D-RoF, digital radio over fibre) such as CPRI (Common Public Radio Interface) [58] or OBSAI (Open Base station Architecture Initiative) [59] for the baseband transmit and receive signals. Figure 27: Step 3: BBU hostelling with stacking based on D-RoF link over optical distribution network, the BBUs are collocated in a central office and are connected to the radio towers through an optical distribution network. Figure 28: Step 4: BBU hostelling with resource pooling (C-RAN), a single baseband unit is connected to a number of radio towers to increase the overall network bandwidth through resource pooling. The next logical step (Figure 27), is to move the BBU to a central office. It is called BBU centralization or BBU hostelling with stacking: BBUs of different base-stations are co-located in the same CO. Generally there is one BBU per antenna site and they communicate with other BBU s within the BBU hostel via standardized X2 interface. No final report - Waiting for acceptance from the European Commission Page 69 of 119

70 Finally, step 4 (Figure 28) is C-RAN or BBU hostelling with resource pooling between all the BBU s in the BBU hostel [60]. Now a centralized set of BBUs with resource pooling is capable of handling a large number of RRHs located at different antenna sites. In steps 3 and 4, thanks to the baseband breakdown, a new connectivity segment called fronthaul appears between the RRHs on the cell site and the BBUs in the central office. The centralization of baseband processing at Central Office extends the reach of digital radio links inter-connecting Remote Radio Unit (RRU) and Baseband Unit (BBU) from the local scope (due to the co-location of BBU/RRU) to a network scope: the consequence is an enhancement of link length from 100 s meter (internal to cell site) to 10 s Km (across a network infrastructure). Due to the hard real-time and capacity requirements, TDM protocols are used for local connectivity and they appear as the most suitable way for realizing also the network wide inter-connection. Two standards, namely Common Public Radio Interface (CPRI) and Open Base Station Architecture Initiative (OBSAI), exist for transport s interfaces between RRH and BBU. CPRI and OBSAI share a number of similarities. In both of them the radio signal is digitised (D-RoF, digital radio over fibre). The predominant deployment of CPRI and the specific interest to this frame format by vendors and operators engaged in defining solutions for BBU centralization make it the preferred option. In terms of fronthaul requirements, important aspects such as bit rate capacity, latency constraints, jitter & synchronization, and fibre availability can be mentioned. Please find more in [61]. The CPRI fronthauling model implies the collection of serial CPRI streams at cell sites, where the sampled antenna signals (I/Q data) related to various mobile technologies (including 2G, 3G, LTE) are mapped into a TDM frame. The result is a constant bit rate signal with line rates in the range Mb/s Gb/s. Current requirements for CPRI interface, in terms of line rate (multi-gb/s) and maximum delay budget associated to the transport network (few hundreds of μs) impact the selection of transport options. Among the technologies currently used for IP backhauling, then, (1) xdsl, (2) GPON/XGPON, (3) microwave transmission, (4) optical transmission (pt-to-pt and WDM), this last one appears as the most suitable option for achieving high rates (10 Gb/s), while keeping low latency. The basic alternatives for CPRI fronthauling are mainly driven by the availability of fibres: thus, the simplest option in a fibre-rich context is based on direct point-to-point fibre connections between RRUs and BBUs across the fronthauling area. An alternative approach, which allows for minimizing the usage of fibres, implies, on contrary, the multiplexing of several CPRI streams over the same physical medium, via WDM (optical), via TDM (electronic) devices or via combination of TDM and WDM. Different topologies for the interconnection RRUs-BBUs may be addressed (point-to-point, tree, ring). In all cases, monitoring and management of these fronthaul networks, including fault, configuration and performance management, are important issues. No final report - Waiting for acceptance from the European Commission Page 70 of 119

71 WDM multiplexing implies that coloured optical modules are equipped within the RRUs and BBUs and multiplexing is performed via a purely passive device: consequently, there is no need of power supply and installation in field is quite easy, but the support of service demarcation function appears more critical than in active solutions (e.g. higher functions integration is needed in the optical modules or external monitoring means are required). CWDM or DWDM grid usage depends on the amount of client signals to be multiplexed and on the topology and population of fronthaul network to be deployed. TDM multiplexing in the access implies that a small electronic box (active) has to be located in the cell site (in place of BBUs), operating as a Cell Site Gateway (CSG). Such a box is reasonably supposed to be temperature hardened in order to not require any air conditioning, to be much more reliable than the BBU stack and to rarely require software upgrades. Possible options for multiplexing function may be: Multiplexing in the CPRI domain, by aggregating into the higher rate of the hierarchy (Rate Gb/s), lower CPRI streams (Rate Mb/s to Rate Mb/s); Encapsulating client CPRIs in Low Order ODU0/flex digital containers, then multiplexing them into one High Order ODU2 digital container (according to ITU-T G.709 OTN hierarchy rules). Encapsulation of client CPRI signals could also be considered in Ethernet. In both cases, latency and jitter requirements have to be satisfied, which needs further developments. TDM-based solutions allow, in principle, to better fill the line interface towards the fronthaul network, but they may show constraints in case of cell site throughput higher than uplink rate (see 10G): in this case either more uplink fibres or, in alternative, a higher rate uplink (e.g. 40G) is required. A combined TDM+WDM solution allows, then, overcoming the scalability issue of pure TDM solutions by minimizing the amount of needed fibres, thanks to the optical multiplexing. Besides, the selection of OTN (as TDM part of the solution) allows for the implicit support of carrier-grade OAM, resilience and management functions. 5.4 Generalized handover and advanced offloading In existing networks, it is typical that a logical single network element is in charge of managing mobility. This impedes routing functionality as all data traffic has to be forwarded to this network element, thus reducing system scalability and overall reliability. When flattening network architectures, new means for supporting advanced offloading and mobility can be envisaged in a more distributed and dynamic fashion. In the following, we introduce the main concepts to be considered in convergent multi-access networks. From a technological perspective, mobility scenarios can be classified in two categories: horizontal and vertical. Horizontal handovers refer to handovers between cells in homogeneous networks of the same technology, while vertical handovers refer to heterogeneous networks and are performed between different radio No final report - Waiting for acceptance from the European Commission Page 71 of 119

72 technologies such as 3G and LTE or 3GPP and non-3gpp technology (Wi-Fi). Unlike horizontal handover, a vertical handover can be performed for convenience (e.g., improving user QoS or offloading traffic from a specific radio access technology) rather than for connectivity reasons. Please note that the definition of heterogeneous network can also have a broader meaning in the context of the RAN deployments. For this case, please see Section Generalized 3D handover This section introduces the description of a generalized handover or 3D handover. Whereas horizontal and vertical handovers refer to an intra-operator scenario, a transversal handover here denotes a handover between the networks of two operators (e.g. in a roaming scenario). This transversal handover can thus be considered as a 3 rd dimension for handover. For horizontal and vertical handovers, standardized mechanisms exist (e.g. as specified by 3GPP in [4]) and are demanded in heterogeneous access networks as described in section 5.2. The regular execution of transversal handover between cells administrated by different operators is required in FMC scenarios involving operator cooperation and share of radio capacity provided by their access networks, such as discussed in section This feature could contribute to cost and energy savings for the operators through sharing of infrastructures and resources, while increasing the bandwidth experienced by the user resulting in higher customer satisfaction. In addition to technical challenges, regulatory and policy issues do apply here predominantly Advanced Offloading The current technology enablers for mobile broadband offloading are the Hot Spot 2.0 initiative of the Wireless Alliance [16], which is based on IEEE standards, and the Access Network Discovery and Selection Function (ANDSF) implementation of 3GPP (3GPP TS Architecture enhancements for non-3gpp accesses ). ANDSF helps the terminal to discover and select other networks (e.g. Wi-Fi) by providing information on the available networks (e.g. if they are trusted or not, network name, security parameters), the network selection policies for the UE s, which can be defined by location (e.g. Cell ID, GPS), date and time, and inter-system routing policy (for example network selection policies for UE s with simultaneous LTE and Wi-Fi, including network selection done by session, flow or service). The ANDSF specification is evolving in 3GPP to overcome some of its limitations, because the ANDSF is a standalone entity than currently does not interface with the PCRF or the HSS, and therefore cannot take into account subscriber s contracted policies. The ANDSF does not interface either with the RAN, and therefore cannot take into account aspects like 3GPP s cell congestions or Wi-Fi s AP capabilities. No final report - Waiting for acceptance from the European Commission Page 72 of 119

73 On the other hand, Hot Spot 2.0 provides automatic discovery, selection and connection with Wi-Fi hotspots, without user intervention, and with secure, encrypted transmissions over the Wi-Fi network. Hot Spot 2.0 release 2 is expected for the end of 2013, and it includes the provisioning of the operator s policy for network selection. However it must be taken into account that Hot Spot 2.0 is currently a Wi-Fi-only technology, as it does not include any inter-working with a 3GPP network. The current trend is towards a convergence of Hot Spot 2.0 and ANDSF solutions, providing full Wi-Fi and 3GPP interworking, including simultaneous and parallel data flows on both radio interfaces, based on operators policies, customer profile and load conditions. Future implementations should consider in particular the load, available data rate and delay of the interfaces between the different network nodes, e.g. the fixed access links to the AP s or base stations, or the links with the ANDSF and the Wi-Fi controllers. To this end, ANDSF future specifications will probably include some of the Hot Spot 2.0 network discovery and automatic connection solutions, along with FMC solutions not implemented yet in any of these standards Distributed Mobility Management Adapting mobility management for evolving from a centralized and hierarchic model to a flat one is a key challenge for the future of mobile networks. This will pave the road toward real convergent networks where main mobility functions (e.g. location management, offloading, access selection, handover, traffic redirections) are activated only when necessary. The main properties to fit to such flat environment are dynamicity and distribution: dynamicity for tuning mobility management support to the requirements of each data flow; distribution for spreading mobility functions and protocols down to the edge of the network and reducing impact on end to end traffic transport and routing. The IETF Distributed Mobility Management (DMM) working group considers requirements [62] and proposals for the distribution and dynamic allocation of mobility anchors closer to the mobile node, down to the access router. Both host based and network based solutions are considered. For example, the basic mobility management operations of DMA, a pioneer proposal for dynamic and Distributed Mobility Anchoring [63][64], are illustrated in Figure 29: at initial attachment on Access Node AN1 (a), the Mobile Node (MN) initiates a traffic flow with any type corresponding host or server, Flow1. AN1 acts as a standard access router and packets are routed towards the IP network using no tunneling or mobility features. In (b), when the MN moves and attaches to AN2, a mobility anchoring context is activated in AN1 for maintaining the delivery of Flow1, through a tunnel towards AN2. Thus, traffic flows created prior to the handover remain delivered to/from the same MN IP address. However, while being attached to AN2 the MN configures a new IP address to be used for any new flow that needs to be set up, like Flow2 shown on (c). When Flow1 is completed, the corresponding anchoring context and tunneling resources in AN1 can be released, see (d). No final report - Waiting for acceptance from the European Commission Page 73 of 119

74 Figure 29: Mobility operation of DMA scheme DMM approaches are very promising in the framework of convergent and multiaccess networks, where different access routers will cooperate to offer on demand mobility support, by activating mobility anchoring and indirections contexts when required. Such new (flat) architecture thinking is made possible with only little evolutions of standard protocols, as shown in the IETF where several proposals consider the use of existing protocols like Mobile IPv6 and Proxy Mobile IPv6. However, there may be more effects on deployment and manageability purposes while interactions with other AAA and QoS need to be carefully studied. No final report - Waiting for acceptance from the European Commission Page 74 of 119

75 6 REFERENCES [1] COMBO deliverable D2.1: Framework reference for fixed and mobile convergence, V1.0, September [2] COMBO deliverable D2.2: Roadmaps for independent fixed and mobile network evolution, V1.0, September [3] Broadband Forum TR101: Migration to Ethernet-Based Broadband Aggregation. [4] 3GPP TS : E-UTRA/E-UTRAN overall description Stage 2 (Release 11, V11.4.0, ). [5] 3GPP TS : General Packet Radio Service (GPRS) enhancements for Evolved Universal Terrestrial Radio Access Network (E-UTRAN) access (Release 11, V11.4.0, ). [6] 3GPP TS : Architecture enhancements for non-3gpp accesses (Release 11, V11.5.0, ). [7] 3GPP TS : Network architecture (Release 12, V12.1.0, ). [8] 3GPP TR : Vocabulary for 3GPP Specifications (Release 11, V11.3.0, ). [9] 3GPP TS : "3GPP System Architecture Evolution: Security Architecture". [10] TR-069 Amendment 4, CPE WAN Management Protocol, Broadband Forum Technical Report. [11] ITU-T G GPON: ONT management and control interface specification. ITU-T recommendation. [12] TR-142 Issue 2, Framework for TR-069 enabled PON Devices, Broadband Forum Technical Report. [13] Broadband Forum Technical Report TR-203 Interworking between Next Generation Fixed and 3GPP Wireless Networks, August [14] Broadband Forum Technical Report TR-134 Broadband Policy Control Framework (BPCF), July [15] Wi-Fi Alliance Hotspot 2.0 Technical Task Group. Wi-Fi CERTIFIED Passpoint (Release 1) Technical Specification Version [16] Wi-Fi Alliance Hotspot 2.0 Technical Task Group. Wi-Fi CERTIFIED Passpoint (Release 1) Deployment Guidelines. Version October [17] 3GPP TS : Access Network Discovery and Selection Function (ANDSF) Management Object (MO). [18] PMIP: RFC 5213 Proxy Mobile IPv6 ( ). [19] PMIP for IPv4: RFC 5844 IPv4 Support for Proxy Mobile IPv6 ( ). No final report - Waiting for acceptance from the European Commission Page 75 of 119

76 [20] PMIP IP flow mobility: draft-ietf-netext-pmipv6-flowmob-07 Proxy Mobile IPv6 Extensions to Support Flow Mobility ( ). [21] PMIP: TS Proxy Mobile IPv6 (PMIPv6) based Mobility and Tunnelling protocols; Stage 3 (Release 12; V12.0.0; ). [22] PMIP/GTP IP flow mobility: TR Network based IP flow mobility (Release 12; V1.7.0 ; ). [23] Cisco White Paper: Cisco Visual Networking Index: Forecast and Methodology, , 29, May [24] Malandrino, F.; Casetti, C.; Chiasserini, C., "Content Discovery and Caching in Mobile Networks with Infrastructure," IEEE Transactions on Computers, vol.61, no.10, pp.1507,1520, Oct [25] Negin Golrezaei, Karthikeyan Shanmugam, Alexandros G. Dimakis, Andreas F. Molisch, and Giuseppe Caire, "FemtoCaching: Wireless video content delivery through distributed caching helpers," Proceedings IEEE INFOCOM, pp , March [26] Shinae Woo, Eunyoung Jeong, Shinjo Park, Jongmin Lee, Sunghwan Ihm, and KyoungSoo Park. Comparison of caching strategies in modern cellular backhaul networks. In Proceeding of the 11th annual international conference on Mobile systems, applications, and services (MobiSys '13). New York, NY, USA, [27] J. Hansryd, J. Edstam, B.-E. Olsson, and C. Larsson, Non-line-of-sight microwave backhaul for small cells, Ericsson review, Feb. 2013, available at 3/er-nlos-microwave-backhaul.pdf. [28] Evolving to Microwave Ring Protection, Alcatel-Lucent Application Note, May 2013, Available at [29] M. Kretschmer, C. Niephaus, and G. Ghinea, Towards QoS Provisioning in a Heterogeneous Carrier-Grade Wireless Mesh Access Networks Using Unidirectional Overlay Cells, LNICST [30] Project website TUCAN3G, [31] A. Ford, C. Raiciu, M. Handley, S. Barre, and J. Iyengar, "Architectural Guidelines for Multipath TCP Development", RFC 6182, March [32] A. Ford, C. Raiciu, M. Handley, and O. Bonaventure, TCP Extensions for Multipath Operation with Multiple Addresses, RFC 6824, January [33] K. Kyu-Han Kim and K.G. Shin, PRISM: Improving the Performance of Inverse- Multiplexed TCP in Wireless Networks, IEEE Transactions on Mobile Computing, (Volume:6, Issue: 12 ), [34] K. Evensen, D. Kaspar, P. Engelstad, A.F. Hansen, C. Griwodz, P. Halvorsen, A network-layer proxy for bandwidth aggregation and reduction of IP packet No final report - Waiting for acceptance from the European Commission Page 76 of 119

77 reordering, IEEE 34th Conference on Local Computer Networks, 2009 (LCN 2009), p [35] D. Zhou, W. Song, M. Shi, Goodput Improvement for Multipath TCP by Congestion Window Adaptation in Multi-Radio Devices, IEEE CCNC (Consumer Communications and Networking Conference) 2013, Las Vegas, USA. [36] IETF WG MultiPath TCP website, [37] E. Cuervo, A. Balasubramanian, Dae-ki Cho, A. Wolman, S. Saroiu, R. Chandra, and P. Bahl, MAUI: Making Smartphones Last Longer with Code offload, in Proceedings of the 8th International Conference on Mobile systems, applications, and services, pp , June [38] Byung-Gon Chun, Sunghwan Ihm, Petros Maniatis, Mayur Naik, and Ashwin Patti CloneCloud: elastic execution between mobile device and cloud. In Proceedings of the sixth conference on Computer systems (EuroSys '11). New York, USA, [39] Mahadev Satyanarayanan, Paramvir Bahl, Ramon Caceres, and Nigel Davies, The Case for VM-based Cloudlets in Mobile Computing, IEEE Pervasive Computing, [40] Tim Verbelen, Pieter Simoens, Filip De Turck, and Bart Dhoedt, Cloudlets: bringing the cloud to the mobile user, the 3rd ACM workshop on Mobile cloud computing and services, pages 29-36, NY, USA, [41] OpenFlow Management and Configuration Protocol (OF-Config 1.1.1). Open Networking Foundation. March 23, [42] Software-Defined Networking: The New Norm for Networks. Open Networking Foundation. April [43] RFC OSPFv [44] NOX: towards an operating system for networks. Natasha Gude & all SIGCOMM Comput. Commun. Rev. 38, 3 (July 2008), [45] Network Functions Virtualisation An Introduction, Benefits, Enablers, Challenges & Call for Action, WP, October [46] The Economics of Cloud Computing Addressing the benefits of infrastructure in the cloud. Ted Alford and Gwen Porton. Booz Allen Hamiton Inc, [47] ETSI GS NFV 001, Network Function Virtualisation, Use Cases, October [48] ETSI GS NFV 002, Network Function Virtualisation, Architecture Framework, October [49] ETSI GS NFV 003, Network Function Virtualisation, Terminology for Main Concepts in NFV, October [50] ETSI GS NFV 004, Network Function Virtualisation, Virtualisation Requirements, October No final report - Waiting for acceptance from the European Commission Page 77 of 119

78 [51] Laraqui, Kim. Small Cell Optical Mobile Backhauling: Architectures, Challenges, and Solutions. ECOC Sept [52] Ericsson white paper Heterogeneous Networks. Networks.pdf. [53] 3GPP TS CR 0001 RP : Correction on Ideal backhaul latency (Release 12, V12.0.0, ). [54] G. Kardaras, J. Soler, L. Brewka, L. Dittmann, Fibre to the antenna: A step towards multimode radio architectures for 4G mobile broadband communications, 2010 IEEE 4th International Symposium on Advanced Networks and Telecommunication Systems (ANTS), pp [55] P. McClusky, J. L. Schroeder, Fibre-to-the-antenna: Benefits and protection requirements, Telecommunications Energy Conference (INTELEC), Sept Oct , pp [56] A. Pizzinat, P. Chanclou, F. Le Clech, B. Landousies, «FTTx where x stays for Antenna: requirements on optical access/distribution network for new mobile backhaul architectures, FTTx SUMMIT Europe, 25th Aptril 2012, London, U-K [57] A. Pizzinat, P. Chanclou, F. Frank, B. Charbonnier, P. Niger, B. Landousies, P. Herbelin, JM. Picard, J-P. Charles, "Infrastructure convergence for fixed and mobile access networks", OFC 2009, Workshop "Migration Scenarios toward Future Access Networks I", San Diego, March 22nd, [58] Common Public Radio Interface (CPRI); Interface Specification, v5.0, 2011 available at [59] [60] China Mobile Research Institute, "C-RAN The Road Towards Green RAN" White Paper Version 2.5 (Oct, 2011), [61] Optical fibre solution for mobile fronthaul to achieve Cloud Radio Access Network Philippe Chanclou, Anna Pizzinat, Fabien Le Clech, To-Linh Reedeker, Yannick Lagadec, Fabienne Saliou, Bertrand Le Guyader, Laurent Guillo, Qian Deniel, Stephane Gosselin, Sy Dat Le, Thierno Diallo, Romain Brenot, Francois Lelarge, Lucia Marazzi, Paola Parolari, Mario Martinelli, Sean O dull, Simon Arega Gebrewold, David Hillerkuss, Juerg Leuthold, Giancarlo Gavioli, Paola Galli- Future Network & Mobile Summit 2013, 3-5 July 2013, Lisbon, Portugal. [62] Chan, H. A. (ed.): Requirements for distributed mobility management. Internet Draft draft-ietf-dmm-requirements (work in progress), Internet Engineering Task Force (2013). [63] Bertin, P., Bonjour, S., Bonnin, J.-M. (2008). A Distributed Dynamic Mobility Management Scheme Designed for Flat IP Architectures. Proceedings of 3rd International Conference on New Technologies, Mobility and Security (NTMS 2008). No final report - Waiting for acceptance from the European Commission Page 78 of 119

79 [64] Distributed Mobility Anchoring, draft IETF draft-seite-dmm-dma-06.txt, P. Seité, P. Bertin JH. Lee, July 2013, work in progress. [65] COMBO internal spreadsheet document, Detailed description of equipment for D3.1 A.2 appendix, October No final report - Waiting for acceptance from the European Commission Page 79 of 119

80 7 CONCLUSION This D3.1 document of COMBO project is the first deliverable of Work Package 3 Fixed Mobile Convergent Architectures. As a main output of Task 3.1 Overall analysis and recommendations, this document prepares the overall work of the work package through a preliminary identification and analysis of network functions, equipment and infrastructures to be implemented in fixed/mobile convergence scenarios. The architecture and roles of key equipment in today s networks were first described with a structural viewpoint, considering the various network types covered by COMBO (fixed, mobile, Wi-Fi). This description showed the main relations between key equipment as well as some protocol stack examples illustrating the actual complexity of current networks. A global picture, called current FMC network, showed in particular how current networks already incorporate some degree of convergence between fixed, Wi-Fi and mobile. So as to prepare future functional analyses of FMC network scenarios, the main network functions implemented in current networks were listed and classified according to eleven functional groups, which were described in details in appendix A.1 of the document. Based on the current network descriptions, the analysis of differences between fixed, Wi-Fi and mobile networks allowed an assessment of the gaps towards network convergence, both from a structural viewpoint and from a functional viewpoint. These gaps towards convergence were related in particular to the technical problems which have to be solved to answer to the use cases derived by WP2. New architectural frameworks (Cloud Computing, Software Defined Networking, Network Function Virtualization, Autonomic Networking) will certainly strongly impact future FMC networks, and are thus important elements to be considered in a top down approach of network convergence. Finally, the activities undertaken up to now have also put to the forefront the following technological concepts, which are believed to be key enablers for network convergence: an Unified optical access & aggregation network; Heterogeneous radio access networks; BBU hostelling and mobile fronthaul technologies; Generalized handover and advanced offloading. The overall work presented in this document will serve as a basis for the definition of some few candidate architectures for FMC networks, which will be milestones of tasks 3.2 and 3.3 of the project. No final report - Waiting for acceptance from the European Commission Page 80 of 119

81 A APPENDIX A.1. Detailed description of functions This part describes each function listed in the functional Groups tables in 2.2. A.1.1 L2 Forwarding: Forwarding Ethernet is the most relevant layer-2 transmission technology in today s access and aggregation networks that increasingly replace the legacy Asynchronous Transfer Mode (ATM) based forwarding. In carrier networks Ethernet forwarding is typically realized on the basis of the bridging technology also known as Ethernet switching. The forwarding decision is made by the network nodes (e.g. Ethernet switches, access node with integrated Ethernet switch) on the basis of the available layer-2 protocol information (e.g. MAC addresses, VLAN-ID). Ethernet switches typically perform MAC address learning in order to generate a correlation between MAC address and destination port. Virtual Local Area Network (VLAN) handling: The VLAN technology is used to split up one Ethernet broadcast domain into multiple distinct broadcast domains which are mutually isolated. The IEEE working group 802.1Q specified the VLAN technology in The specification defines a VLAN tag (4 byte) that can be added to the Ethernet frame header. This tag includes a 12 bit VLAN-Identifier (VID) which allows to address up to 4096 VLANs (note: the IDs 0 and 4095 are reserved). In today s Ethernet based access and aggregation networks the VLAN technology is widely used for separating the traffic of different customers and service providers. In addition it is used for marking the class of service in order to enable traffic prioritization and QoS (please refer toa.1.7). Different VLAN forwarding schemes have been defined by the Broadband Forum in TR-101 and TR-156. The 1:1 VLAN scheme relies on the unique one-to-one binding between user port and a VLAN. According to N:1 scheme, the Access Node is considered to be a VLAN aware bridge, where each N:1 VLAN is a separate Virtual Bridge (VB) instance. This means that each VB performs independent source MAC address learning and frame forwarding process. IEEE 802.1AD Provider Bridges (Q-in-Q): The Q-in-Q concept is an extension of the VLAN technology in order to overcome the limitation of a maximum number of 4094 VLANs. This technology enables Ethernet frames with two VLAN tags (customer C-VLAN and service S-VLAN). No final report - Waiting for acceptance from the European Commission Page 81 of 119

82 IEEE 802.1ah: Provider Backbone Bridging (MAC-in-MAC): MAC-in-MAC is another standardized concept to overcome some limitations of the Ethernet technology, e.g. VLAN scalability. It facilitates upstream core switches, by hiding customer MAC addresses behind fewer backbone MAC addresses. MAC-in- MAC is an encapsulation method that allows encapsulating a complete Ethernet frame including Q-in-Q VLAN tag (as payload) by adding a MAC address to the front of the packet. The MAC header can also support a VLAN tag for the backbone network. L3 Forwarding: Layer-3 forwarding means that a network node is terminating the layer-2 protocols (e.g. Ethernet) at the ingress interface and making the forwarding decision on the basis of layer-3 protocol information (mainly IP) in order to forward the data to the appropriate egress interface. The layer-3 payload is encapsulated in a new layer-2 frame at the egress interface using the MAC address of the network node as source MAC address. IP Routing: IP routing is a method to determinate the path of IP packets across multiple IP networks. IP routing is often realized on the basis of routing protocols. Two major classes of routing protocols can be distinguished: Interior Gateway Protocols (IGPs) exchange routing information within a single Autonomous Systems (AS). An AS is a group of routers that are under the control of a single administration (e.g. network provider) and exchange routing information using a common routing protocol. Examples of IGPs are the Open Shortest Path First (OSPF) and the Intermediate System to Intermediate System (IS-IS) Exterior Gateway Protocols (EGPs) are used for exchanging IP data between Autonomous Systems (AS). The most relevant EGP is the Border Gateway Protocol (BGP) IPv6 support: Due to the limitations of the internet protocol version 4 (IPv4) addressing scheme it is widely expected that the Internet and its applications will migrate to internet protocol version 6 (IPv6). Therefore, an advanced access and aggregation architecture should support a transition to IPv6 while maintaining connectivity for users to the IPv4 Internet as well. This requires transitional technologies that are able to address the needs of both protocols since IPv4 and IPv6 are not interoperable. No final report - Waiting for acceptance from the European Commission Page 82 of 119

83 Network Address Translation (NAT): NAT is a technology that allows a small number of public IP address to be shared by a large number of hosts using private IP addresses. This technology is also widely used in today s broadband networks in order to solve the problem of the limited IPv4 addressing scheme while the IPv6 protocol is deployed. For example, NAT is typically used within the residential gateway (RGW) in order to enable one public IP address to be shared by several CPE devices (PC, Set-top box). Multicast: In computer networking, multicast is the delivery of a message or information to a group of destination computers simultaneously in a single transmission from the source. Copies are automatically created in other network elements, such as routers, but only when the topology of the network requires it. Multicast transmission technology is available at both the data link layer (Layer 2) and the network layer (Layer 3). Multicast offers an efficient alternative to unicast and broadcast transmission as it helps to economize the available bandwidth by minimizing the packet replication in the network. Multicast is most commonly implemented in IP multicast, which is often employed in Internet Protocol (IP) applications of streaming media and Internet television. In IP multicast the implementation of the multicast concept occurs at the IP routing level, where routers create optimal distribution paths for datagrams sent to a multicast destination address. L3 Multicast: L3 multicast means in this context IP multicast which aims at the transmission of an IP data packet to a host group that is defined by a single IP address. This technology is described in the IETF RFC The Internet Group Management Protocol (IGMP) provides a method to automatically control the multicast traffic through the network. The IGMP protocol, version 1, is specified in RFC The IGMP, version 2, is defined in RFC 2236 and the version 3 is specified in RFC L2 Multicast: By default, an Ethernet switch forwards multicast traffic to all ports that belong to the destination LAN. If the listener pool needs to be restricted to specific listeners, several solutions can be used, e.g. IGMP snooping, GARP Multicast Registration Protocol (GMRP). IGMP snooping is the most relevant method that is defined in RFC4541. No final report - Waiting for acceptance from the European Commission Page 83 of 119

84 Multiprotocol Label Switching (MPLS): In MPLS networking, a label-switched path (LSP) is a path through an MPLS network, set up by a signalling protocol such as LDP, RSVP-TE, BGP or CR-LDP. Routers between Provider Edges (PE), which need only swap labels, are called transit routers or label switch routers (LSRs). PE MPLS (Multiprotocol Label Switching): The MPLS PE (Provider Edge) router or network element is both the interface between the customer-facing network and the MPLS core and the point where the data is given an MPLS label and/or the label is removed. Seamless MPLS: S-MPLS is used as a generic term to describe an architecture which integrates access, aggregation and core network in a single MPLS domain (cf. seamless MPLS architecture, draft-ietf-mpls-seamless-mpls). Seamless MPLS connects all MPLS domains on the MPLS transport layer providing a single transport layer for all services - independent of the service itself. The separation of the service and transport plane is one of the key elements. Point-to-Point Protocol (PPP): The purpose of the Point-to-Point Protocol is to initialize point-to-point connections, to keep them up and to determine them again. PPP is a layer-2 protocol that is used for the transmission of layer-3 protocols through a point-to-point connection. In addition, it is used for authentication, network layer auto-configuration and data encryption. Today, PPP is mainly used for configuring the IP connection of the residential highspeed internet access service in broadband networks. Layer-2-Tunneling-Protocol (L2TP): The Layer-2-Tunneling-Protocol is used for tunneling of layer-2 protocols (mainly PPP) through an IP network. L2TP is often used for wholesale applications (e.g. IP- BSA) in today s broadband networks. Generic Routing Encapsulation (GRE) and Proxy Mobile IP (PMIP): GRE is an IP based tunnelling protocol while PMIP is an IP based mobility protocol. Typical usage: the S5 interface connects the LTE serving gateway (S-GW) to the packet data network gateway (P-GW) and is especially required for S-GW reallocation in case a mobile device moves into the service area of a different S-GW. No final report - Waiting for acceptance from the European Commission Page 84 of 119

85 The S5 interface can be based on GTP-tunnelling (see below) or on PMIP and GREtunnelling. GPRS Tunneling Protocol (GTP): GTP is an UDP/TCP based protocol to transport packet data over 3GPP based mobile radio networks (GSM, UMTS, LTE). It is divided into two parts: GTP control plane (GTP-C) is used to transport control data within the network, e.g. control messages from MME to S-GW to establish a new session for a mobile device. GTP user plane (GTP-U) is used to transport user data within the network. The tunnelling functionality is used to support mobility: In case a mobile device moves e.g. in LTE from one enodeb to another, the GTP-U tunnel is redirected to towards the new enodeb. Traffic Offloading: Mobile data offloading is the use of complementary network technologies for delivering data originally targeted for cellular networks. Rules triggering the mobile offloading action can be set by either an end-user (mobile subscriber) or an operator. The code operating on the rules resides in an end-user device, in a server, or is divided between the two. End users do data offloading for data service cost control and the availability of higher bandwidth. Operators do it to ease congestion of cellular networks. The main complementary network technologies used for mobile data offloading are Wi-Fi, femtocell and Integrated Mobile Broadcast (imb). HTTP traffic redirect: URL redirection, also called URL forwarding, is a World Wide Web technique for making a web page available under more than one URL address. When a web browser attempts to open a URL that has been redirected, a page with a different URL is opened. URL redirection can be used for URL shortening, to prevent broken links when web pages are moved, to allow multiple domain names belonging to the same owner to refer to a single web site, to guide navigation into and out of a website, for privacy protection, and for less innocuous purposes such as phishing attacks. A.1.2 Automatic configuration and management Physical medium resources management (control plane): Control of transmission characteristics of a given physical medium. Resources management involves strategies and algorithms for controlling transmission parameters specific to the considered medium and allocating transmission resources to the connected devices. No final report - Waiting for acceptance from the European Commission Page 85 of 119

86 Radio resources management (control plane): Control of radio transmission characteristics in wireless communication systems, such as transmit power, channel allocation, data rates, handover criteria, modulation scheme, error coding scheme, etc. Device management (management plane and control plane): Protocols and mechanisms that achieve management of devices and applications running on them. Management applies (but is not limited) to mobiles, smartphones, Machine-to-Machine (M2M) equipment and in general any device capable to connect to data network; it includes (but is not limited to): setting, installation and management of initial and operational configuration information related to device capabilities and applications functionalities, firmware and software update, retrieval of management information from devices, processing events and alarms generated by devices, running diagnostic tests and monitoring tasks, controlling device capabilities and applications, controlling and managing how applications running on the devices uses and interacts with underlying capabilities. Logical Link management (VLANs ) (management plane): Protocol and mechanisms that achieve the segmentation of a given link layer into several logical links, isolated from each other. The segmentation is usually realized on switches and router devices and it allows sharing layer 2 infrastructures (links, switches) among several IP networks and routers. IP infrastructure management (management plane): Protocols and mechanisms that achieve management of a given IP infrastructure (routers, gateways ). It includes IP (v4 and or v6) addressing and sub-networks configuration, routing protocols activation, routers features and parameters configuration, firewalling rules. Customer IP pools management (management plane): Protocols and mechanisms that achieve the allocation of one or several pools of IP addresses to a given customer. Subscriber database (management plane and control plane): User database containing subscription related information in subscriber profiles. The subscriber database is used to perform authentication and authorization of the end No final report - Waiting for acceptance from the European Commission Page 86 of 119

87 user, it can provide information about the subscriber's location, IP information, services access rights, QoS and user s traffic related policy information. SNMP (management plane and control plane): Simple Network Management Protocol (SNMP) allows to manage devices in IP networks. SNMP operates on layer 7 of internet protocol suite. SNMP agents are run on managed devices so as to expose management data to the managers. SNMP can be used to monitor IP devices for specific conditions. It also allows active management tasks, e.g. pushing a new configuration through remote modification of some variables. A.1.3 Resilience Redundancy: Redundancy, in a network environment, is a method that ensures that connectivity continues to function in presence of faults, thanks to redundant interconnection or redundant parts of network elements, eliminating network downtime caused typically by a single point of failure. Redundancy functions are included in all OSI layers, for example, at physical level using dual power supply units (PSUs), ITU-T Recommendations G.8031 and G.8032 to protect Ethernet linear/ring switching respectively, Virtual Router Redundancy Protocol (VRRP, IETF RFC 3768) or Bidirectional Forwarding Detection (BFD, IETF RFC 5880) at layer 3 to detect faults between 2 endpoints, Multipath TCP (IETF RFC 6824) or checksum as error detection mechanisms at level 4, Real Time Control Protocol (RTCP, IETF RFC 3550) to provide statistics for a RTP flow at level 5, and probes for services or applications running at level 6 (e.g. SSL) and 7 (DNS, HTTP, FTP, etc.). Specific functionalities included under this section are: Performance monitoring: Performance monitoring includes the supervision and measurement of relevant performance metrics to assess the performance of a service or a network. Performance monitoring in a redundancy environment is used to detect anomalous situations. Fault detection: Fault defection is the ability of a system of identifying fault conditions based on specific measurements and conditions and issue different alarms depending on the fault type. Network switching: No final report - Waiting for acceptance from the European Commission Page 87 of 119

88 Network switching, under the scope of redundancy, activates the backup links, resources or other mechanisms under fault situations, forwarding the current request to the backup system. Trouble reporting: This includes resolution of some types of problems. As redundancy is related to fault management, trouble reporting is needed to track problems to restore the normal service operation and communicate failures that could affect the service or the network availability (e.g. SLA issues). Normal service restoration: Normal service restoration refers to the ability of a service or a network to restore or recover from a failure. That typically implies to restore the main operational servers or links in order to free backup resources. Load balancing: Load balancing is the ability to balance traffic across two or more communications links or network elements in order to provide network resilience and/or scalability. Specific functionalities included under this section are: Queue management: Queue management in a load balancing environment includes the algorithms to manage the ingress and egress queues policies in the most efficient way and avoid the system congestion. Performance monitoring: Performance monitoring for load balancing includes the supervision and measurement of relevant performance metrics to assess the performance of the load balancing system and provide information to the queue manager that can be used to change queues policies. Forwarding requests: Forwarding requests refers to the ability of a load balancing system to forward initial requests (transparently from the client point of view) to the final server or systems that will process them. Security point (e.g.: against DDoS): No final report - Waiting for acceptance from the European Commission Page 88 of 119

89 Load balancing systems typically provide the front-end to the users, hiding the internal network to the clients, so some security related functionality is needed in order to prevent attacks or intrusions. Session persistence (multiple requests from the same user): Session persistence is a feature of a load balancing system that forces multiple requests from a single client in a single session to be directed to the same backend server. Caching: Caching refers to the ability of a load balancing system to store static content and handle requests without contacting the backend servers. A.1.4 Security This functional group includes the functionalities related to network security. These can be roughly divided into two groups. The first includes the functionalities for the protection of communications either between the user and the Access Server or between the operator s network nodes; The second group includes functionalities aiming at preventing and monitoring unauthorized access and misuse of the network resources. Specific Functionalities in the first group include: End-user authentication: In wired networks, the network operator controls the access points and can link user identity to an access point and use that information in authenticating users. In wired networks, a common scenario is an authentication challenge over Point-to-Point Protocol (PPP) using Password Authentication Protocol (PAP), Challenge- Handshake Authentication Protocol (CHAP), or Extensible Authentication Protocol (EAP), depending on whether the authentication is performed over a secure channel or over an insecure domain such as a Wi-Fi hotspot. Another approach is to rely on DHCP relay with option 82 (RFC 3046). In wireless networks, the presence of a communication is not easily linked to the user who can be nomadic or mobile, therefore requiring cryptographic techniques for authentication. Those mechanisms can be based on passwords, X.509 certificates, or keys stored in tamperproof devices such as SIM Cards. In mobile networks, Authentication and Key Agreement No final report - Waiting for acceptance from the European Commission Page 89 of 119

90 (AKA) allows the network to verify the identity of the mobile user and also to derive a session key for further encryption. Session key agreement and cryptographic tunnel management: In wired networks, eavesdropping of data flows within the network is usually not possible, except by the network operator. Traffic encryption is not mandatory, although it can be used. IPSec Encapsulating Security Payload (ESP) tunnels are a common choice and endpoint authentication and session key generation uses Internet Key Exchange (IKE) with preshared keys or certificates. In insecure domains, such as wireless networks, eavesdropping and data forging or tampering is easy, making encryption necessary, generally by using IPSec, TLS, or the UMTS and LTE encryption algorithms. In the case of mobile networks, the AKA procedure is also used to perform session key agreement. Specific Functionalities in the second group include: Firewalls and layer 2 security: Networks are generally subject to attacks from malicious sources, both from the user s side and from other networks. Given the wide variety of attacks, it is necessary to implement a diverse range of techniques. Access network devices generally implement consistency checks to block forged packets such as Dynamic ARP Inspection, ARP-Reply-Agent, MAC-Address Limitation, MAC Anti-Spoofing. Access Control to network resources and user s data: Usage of network resources is generally regulated and must be enforced. Access networks generally implement techniques to prevent uncontrolled usage of resources, such as rate limiting for broadcast- and multicast frames, limitation of packet flooding and blocking of packets among hosts inside the same network. Further, depending on the network policy and on subscription, some services can be banned, therefore the network must identify and block attempts to use blocked services. A.1.5 OAM & Management Operations, administration, maintenance and management describes the processes, activities, tools and standards involved with operating, administering, managing and maintaining a system. The logical model Telecommunications Management Network (TMN) provides a general, structured view of OAM and Management related to networks, services and No final report - Waiting for acceptance from the European Commission Page 90 of 119

91 others. It is described in recommendation by ITU-T: M.3010: Principles for a telecommunications management network. FCAPS is the ISO Telecommunications Management Network model and framework for network management. FCAPS stands for fault, configuration, accounting/administration, performance and security. FCAPS is refined as part of M.3400: TMN management functions. FCAPS can be seen as the predecessor of the newer FAB (fulfilment, assurance, billing) model defined in ETOM (Enhanced Telecom Operations Map), published by the TM Forum. The FCAPS model can be seen as bottom-up or network-centric. The FAB model looks at the processes more from top-down, is customer/business-centric. The elements of the logical model are depicted in the following figure. Figure 30: OAM & Management model The above model can be viewed by two orthogonal aspects: Vertically from physical entities through the network and services to business management. The other aspect includes management tasks: fault, configuration, accounting, performance and security management. Business management includes high level design, financial plans and control, declaration of aims, decision making, and business level agreements. No final report - Waiting for acceptance from the European Commission Page 91 of 119

92 Service management incorporates customer support: starting and modifying services, accounting tasks, performance management, fault management. Service level agreements are ensured by based on information received from the network and from the customers. Network management includes global management of the network or its segments. Uses the information received from lower layers. It provides framework for services. Element management includes management of network elements and the physical infrastructure. It is typically performed by the infrastructure management staff. Main relevant network OAM tools are given and described hereafter: Ethernet: ITU-T Y.1731: Supported functions: Continuity Check, Connectivity Verification, Alarm suppression, Locked Indication, Remote Defect Indication, Client Signal Fail, Performance (Frame Loss, Frame Delay, Frame Delay Variation, Throughput), Fault localization. IEEE 802.1ag: Supported functions: Continuity Check, Connectivity Verification, Alarm suppression, Remote Defect Indication, Fault localization. IEEE 802.3ah (Link Ethernet First Mile OAM): Supported functions: Continuity Check, Connectivity Verification, Remote Defect Indication, Performance (Errored Symbol Period, Errored Frame Event, Errored Frame Period Event, Errored Frame Seconds Summary Event), Fault localization, Dying Gasp. MPLS: BFD: Supported functions: Continuity Check, Remote Defect Indication. LSP Ping/Traceroute: No final report - Waiting for acceptance from the European Commission Page 92 of 119

93 Supported functions: Continuity Check, Connectivity Verification, Remote Defect Indication, Fault localization, Path tracing. VCCV: Supported functions: Continuity Check, Connectivity Verification, Remote Defect Indication, Fault localization, Path tracing. MPLS-TP: ITU-T G (based on IETF draft-bhh) IETF RFC 6375, RFC 6426, RFC 6427, RFC 6428, RFC 6435 (base for ITU-T G ): Supported functions: Continuity Check, Connectivity Verification, Alarm suppression, Locked Indication, Remote Defect Indication, Client Signal Fail, Performance (Frame Loss, Frame Delay, Frame Delay Variation, Throughput), Fault localization. IP: TWAMP (IP SLA): Supported functions: Performance (Packet Loss, Packet Delay, Packet Delay Variation, MTU Size Changes) SDP Ping/Traceroute: Supported functions: Continuity Check, Connectivity Verification, Remote Defect Indication, Fault localization, Path tracing OTN: ITU-T G.709/G.798: Supported functions: Continuity Check, Connectivity Verification, Alarm suppression, Remote Defect Indication, Client Signal Fail, Performance (BBE, ES, SES, UAS, Delay coarse ODUk round trip). WDM: OTDR (Optical Time Domain Reflectometer): Supported functions: Continuity check, Fault localization No final report - Waiting for acceptance from the European Commission Page 93 of 119

94 A.1.6 Synchronization Telecom providers, mobile operators and private network owners everywhere are migrating to NGN (next-generation networks) architecture to benefit from packet transport s ultra-high throughput and substantial cost savings. Especially mobile base-stations need a highly accurate timing signal that has to be shared across the network. If an individual base-station drifts outside of the specified +/-50 PPB (Parts per Billion) limit, mobile handoff performance degrades resulting in high disconnected calls rate and poor data services quality. As long as mobile base-stations rely purely on TDM based T1/E1 or SONET/SDH backhaul connections, synchronization is not an issue. Yet as the aggregate cost of TDM backhaul connections rises, driven by the need to deliver more and more capacity in support of data and video applications, operators are beginning to transfer their networks to more cost-efficient packet-based synchronization solutions. This move breaks the end-to-end clock synchronization chain that enabled 2G and early 3G networks to keep synchronized. However, while clocking data has been transmitted natively in TDM networks, new Packet Switched Networks (PSNs) are asynchronous by nature and introduce inaccuracies, such as packet delay variation (PDV) and packet loss making synchronization the biggest challenge in the migration to Ethernet, IP and MPLS transport. Unlike legacy TDM networks, packet-based networks are not deterministic. Packets may follow more than one route from source to destination and their order of arrival is not necessarily similar to the order in which they are transmitted. Packets may also be lost on the way and require retransmission. It s plain to see then, that packet based networks will require a specialized device in order to support the submicrosecond levels of synchronous timing and frequency needed by mobile basestations. The evolution of synchronization over packet solutions has involved several approaches, resulting in the development of various methods and standards, the most notable of which include the following: Adaptive Clock Recovery (ACR): ACR is a method in which the clock is distributed over the PSN as an in-band TDM stream and regenerated using the packets time-of-arrival information, independently of the physical layer. The clock stream format is a standard pseudo-wire flow, simplifying interoperability with third-party equipment. In addition, bandwidth consumption can be minimized by using a multicast pseudo-wire for clock distribution. Today, pseudo-wire gateways incorporating high-performance adaptive clock recovery mechanisms are already deployed and meet stringent GSM/UMTS requirements. IEEE (1588v2): No final report - Waiting for acceptance from the European Commission Page 94 of 119

95 IEEE 1588, also known as Precision Time Protocol, is a frequency and time of day distribution protocol, which is based on timestamp information exchange in a masterslave hierarchy, whereby the timing information is originated at a Grandmaster clock function that is usually traceable to a Primary Reference Clock (PRC) or Coordinated Universal Time (UTC). Similar to NTP (Network Time Protocol), it nonetheless offers better accuracy, with HW-based time-stamping support and fractional nanosecond precision. IEEE 1588 defines the packet format for timing distribution but does not specify the actual clock recovery algorithm the critical element in network synchronization. Although it can be implemented end-to-end, support of 1588 by intermediate network elements ( boundary clocks and transparent clocks ) ensures better performance. Network Time Protocol (NTP): NTP is a widely deployed IETF standard for distributing time of day in wide area networks (usually for public Internet). It uses a hierarchical system of "clock strata", whereby the stratum levels define the distance from the master reference clock and, consequently, the associated accuracy. NTPv4 is an enhancement of the NTP protocol, providing an accuracy level of milliseconds. In order to achieve the exacting accuracy required for real-time services, a high-precision (and costly) oscillator is required at the customer premises, or base stations. Synchronous Ethernet (Sync-E): Synchronous Ethernet, defined in ITU-T standards G.8261, G.8262 and G.8264, uses the Ethernet physical layer to accurately distribute frequency, using clock mechanisms similar to those of SDH/SONET. Unlike timing distribution in emulation services, where clocking information is carried in the same flow as the data payload, in Synchronous Ethernet the bits clock of the Ethernet physical layer is disciplined to a PRC, regardless of the higher layer transmission protocols used. As SyncE is a link-by-link frequency distribution scheme, it requires the entire clock distribution path (i.e. all the network nodes involved) to be Sync-E compliant. Network Timing Reference (NTR): NTR is a highly accurate standardized method for frequency distribution in DSLbased Last Mile segments. A network reference clock (i.e. a service clock) is distributed from the DSLAM to the CPE by mapping its clock information to the DSL modem transmission. Depending on the specific DSL technology, this is achieved by either directly locking the DSL symbol clock to the reference clock or by mapping to the DSL frame phase difference bits information between the reference clock and the DSL free-running symbol clock. The advantages of NTR lie in its high level of accuracy and in the fact that it eliminates the need for advanced synchronization hardware in the DSL modem or Integrated Access Device, thereby reducing the overall cost of the solution. No final report - Waiting for acceptance from the European Commission Page 95 of 119

96 Other methods for synchronization over packet include GPS and hybrid topologies involving a separate E1/T1 link for synchronization purposes. Next generation networks, however, are most likely to utilize some sort of combination between several methodologies. For example, NTR can be used between a DSL gateway and the local DSLAM, which can employ ACR, SyncE or 1588v2 towards the packet network. GSM, WCDMA, LTE-FDD and Femto Cells only require frequency synchronization. This is optimally provided by the backhauling networks with Synchronous Ethernet which provides SDH equivalent synchronization by deriving the timing from the physical links. Synchronization is handed over to the relevant cells by the cell site transport gateway via E1 interfaces or BITS interfaces at 2MHz or 2Mbps signals, all deriving the frequency from the incoming SyncE signals. The key drawback of SyncE is the need to implement it all across the network. Operators sometime use older packet switches that do not support SyncE. In these cases, implementation of IEEE1588v2 can overcome this limitation. In other words, SyncE is used wherever possible and IEEE1588v2 can be used to overcome the network sections which do not provide SyncE. Other technologies such as LTE-TDD require, in addition to accurate frequency synchronization, a high level of phase synchronization, or even time synchronization. As SyncE is not capable of providing phase sync, operators can deploy the IEEE1588v2 technology. Moreover, some of the LTE-TDD equipment offers embedded support for 1588v2. In such cases, the backhauling network should provide means of maintaining clock accuracy by means of boundary and transparent clock. For practical reasons, mobile operators deploy new technologies in existing cell sites. The result is that over 60% of mobile cell sites are equipped with a mixture of technologies. Furthermore, every new deployment of mobile backhauling networks takes into account the future need for synchronization. A.1.7 Policy Control: Policy and Charging The process whereby the Policy and Charging Rules Function (PCRF) indicates to the Policy and Charging Enforcement Function (PCEF) how to control the IP- Connectivity Access Network (CAN) bearer. Policy control comprises functionalities for: Gating control, i.e. the blocking or allowing of packets, belonging to a service data flow, to pass through to the desired endpoint; Event reporting, i.e. the notification of and reaction to application events to trigger new behaviour in the user plane as well as the reporting of events related to the resources in the GW (PCEF); QoS control, i.e. the authorisation and enforcement of the maximum QoS that is authorised for a service data flow or an IP-CAN bearer. No final report - Waiting for acceptance from the European Commission Page 96 of 119

97 IP-CAN bearer establishment for IP-CANs that support network initiated procedures for IP-CAN bearer establishment. Gate Enforcement: Along with QoS Enforcement one of the way the PCEF enforces the Policy control: a service data flow, which is subject to policy control, is allowed to pass through if and only if the corresponding gate is open. QoS definition/authorization: The process of defining and authorizing QoS parameters (i.e. QCI, GBR, MBR and the data rate) for a service data flow. Quality of service is the ability to provide different priorities to different applications, users, or data flows, or to guarantee a certain level of performance to a data flow. For example, a required bit rate, delay, jitter, packet dropping probability and/or bit error rate may be guaranteed. Quality of service guarantees are important if the network capacity is insufficient, especially for real-time streaming multimedia applications such as voice over IP, online games and IP-TV, since these often require fixed bit rate and are delay sensitive, and in networks where the capacity is a limited resource, for example in cellular data communication. QoS rules: A set of information enabling the detection of a service data flow and for performing bearer binding and uplink bearer binding verification. The QoS rules contain QoS parameters. QoS enforcement: Along with Gate Enforcement one of the way the PCEF enforces the Policy control. Possible flavours include: QoS class identifier correspondence with IP-CAN specific QoS attributes. The PCEF shall be able to convert a QoS class identifier value to IP-CAN specific QoS attribute values and determine the QoS class identifier value from a set of IP-CAN specific QoS attribute values. PCC rule QoS enforcement. The PCEF shall enforce the authorized QoS of a service data flow according to the active PCC rule (e.g. to enforce uplink DSCP marking). IP-CAN bearer QoS enforcement. The PCEF controls the QoS that is provided to a combined set of service data flows. The policy enforcement function ensures that the resources which can be used by an authorized set of service No final report - Waiting for acceptance from the European Commission Page 97 of 119

98 data flows are within the "authorized resources" specified via the Gx interface by "authorized QoS". The authorized QoS provides an upper bound on the resources that can be reserved (GBR) or allocated (MBR) for the IP-CAN bearer. The authorized QoS information is mapped by the PCEF to IP-CAN specific QoS attributes. Charging Control: The process of associating packets, belonging to a service data flow, to a charging key and applying online charging and/or offline charging, as appropriate, for rating purposes. The PCC charging shall support the following charging models: Volume based charging; Time based charging; Volume and time based charging; Event based charging; No charging (i.e. charging control is not applicable). Policy & Charging Control decision: A PCC decision consists of a set of information (PCC rules) enabling the detection of a service data flow and providing parameters for policy control and/or charging control. Policies can include charging control rules and parameters, traffic flow QoS parameters, traffic forwarding rules, blocking or allowing packets PCC architecture & functions: The PCC functionality includes the following functions: PCRF: Policy and Charging Rules Function is the function devoted to determine real-time policy rules in a multimedia network: the PCRF decides how a certain service data flow shall be treated in the PCEF, and ensure that the PCEF user plane traffic mapping and treatment is in accordance with the user's subscription profile. The PCRF is a software component that operates at the network core and accesses subscriber databases and other specialized functions, such as a charging system, in a centralized manner. PCEF: Policy and Charging Enforcement Function is the function encompassing service data flow detection, policy enforcement and flow based charging functionalities. This functional entity is located at the Gateway (e.g. GGSN in the GPRS case, and PDG in the WLAN case). It provides service data flow detection, user plane traffic handling, triggering control plane session management (where the IPCAN permits), QoS handling, and service data flow measurement as well as online and offline charging interactions. A PCEF shall ensure that an IP packet, which is discarded at the PCEF as a result from No final report - Waiting for acceptance from the European Commission Page 98 of 119

99 policy enforcement or flow based charging, is neither reported for offline charging nor cause credit consumption for online charging. AF: Application Function is an element offering applications that require dynamic policy and/or charging control over the IP-CAN user plane behaviour. The AF shall communicate with the PCRF to transfer dynamic session information, required for PCRF decisions as well as to receive IP-CAN specific information and notifications about IPCAN bearer level events. An AF may communicate with multiple PCRFs. SPR: Subscription Profile Repository is a logical entity containing all subscriber/subscription related information needed for subscription-based policies and IP-CAN bearer level PCC rules by the PCRF. OFCS: Offline Charging System is a process where charging information for network resource usage is collected concurrently with that resource usage and processed in such a way to result in CDR files (Charging Data Record) transferred to the network operator's Billing Domain for the purpose of subscriber billing and/or inter-operator accounting (or additional functions, e.g. statistics, at the operator s discretion). OFCS is a mechanism where charging information does not affect, in real-time, the service rendered; but, it may trigger the PCEF to initiate an IP-CAN bearer service termination at any point in time. OCS: Online Charging System is a process where charging information for network resource usage is collected concurrently with that resource usage in the same fashion as in offline charging. The authorization for the network resource usage is negotiated between the network and the OCS prior to the actual resource usage to occur. OCS is a mechanism where charging information can affect, in real-time, the service rendered and therefore a direct interaction of the charging mechanism with the control of network resource usage is required. BBERF: Bearer Binding and Event Reporting Function includes the bearer binding function (i.e. the function devoted to associate a service data flow to an IP-CAN bearer transporting that service data flow), the uplink bearer binding verification and the event reporting to the PCRF. The PCC architecture includes the following reference points: Rx: the Rx reference point resides between the AF and the PCRF. It enables transport of application level session information from AF to PCRF and it includes (not limited to): IP filter information to identify the service data flow for policy control and/or differentiated charging; Media/application bandwidth requirements for QoS control. The Rx reference point enables the AF subscription to notifications on signalling path status of AF session in the IPCAN. Gx: the Gx reference point resides between the PCEF and the PCRF. It enables a PCRF to have dynamic control over the PCC behaviour at a PCEF and it supports the following functions: request for PCC decision from PCEF to No final report - Waiting for acceptance from the European Commission Page 99 of 119

100 PCRF; provision of PCC decision from PCRF to PCEF; Negotiation of IP-CAN bearer establishment mode (UE-only, UE/NW or NW-only); termination of Gx session (corresponding to an IP-CAN session) by PCEF or PCRF. Sp: the Sp reference point lies between the SPR and the PCRF. The Sp reference point allows the PCRF to request subscription information related to the IP-CAN transport level policies from the SPR based on a subscriber ID, a PDN identifier and possible further IP-CAN session attributes. Gy: the Gy reference point resides between the OCS and the PCEF. It allows online credit control for service data flow based charging. Gz: the Gz reference point resides between the PCEF and the OFCS. It enables transport of service data flow based offline charging information. S9: The S9 reference point resides between a PCRF in the HPLMN (H-PCRF) and a PCRF in the VPLMN (V-PCRF). For roaming with PCEF in visited network, the S9 reference point enables the Home PCRF to have dynamic control, via the V-PCRF, over the PCC behaviour at a PCEF in the VPLMN. In all roaming scenarios, S9 has functionality to provide dynamic QoS control policies from the HPLMN, via a V-PCRF, to a BBERF in the VPLMN. Network authorization: The process granted by the Online Charging System upon request from the network for the network resource usage. When receiving a network resource usage request, the network assembles the relevant charging information and generates a charging event towards the OCS in real-time. The OCS then returns an appropriate resource usage authorization. The resource usage authorization may be limited in its scope (e.g. volume of data or duration), therefore the authorization may have to be renewed from time to time as long as the user s network resource usage persists. Packet policing related functions: Classification: The function of packet policing aiming to identify incoming packets based on specific parameter(s), for internal processing as forwarding, filtering and, where applicable learning). Depending on the data technology, classification may be performed: on port, MAC Address (Source and Destination), VLAN tag, VLAN Priority bits (i.e. Class of Service bits) base, when Ethernet flows are considered; on port, IP Address (Source and Destination), Protocol field, TOS (Type of Service) and DiffServ (Differentiated Service) base, when IP flows are considered; on MPLS label, Traffic Class bits base, when MPLS flows are considered. No final report - Waiting for acceptance from the European Commission Page 100 of 119

101 Metering: The function of packet policing aiming to measure the rate and the burst size of identified incoming packets with respect to pre-defined (and provisioned) parameters. Depending on the policing modes and applications, these parameters include maximum allowed rate (e.g. PIR, MBR, EIR), guaranteed rate (e.g. CIR, GBR), burst size (e.g. PBS, EBS). The metering function may be performed considering also a marker value (a.k.a. colour ) embedded in the incoming packet by the packet source for assigning a specific traffic/bandwidth profile; depending on the technology, colour information is carried by VLAN Priority bits, TOS/DiffServ bits or EXP bits. The enabling of colour processing is done on the base of provisioned value of Colour Mode (CM) parameter (blind or aware). Marking: After metering, packets are marked (or re-marked) according to policy locally handled, which generically includes: packets eligible for the forwarding (with different priorities of queuing) and packets not matching the metering parameters provisioned and, thus, to be discarded (dropped). Scheduling: The function of packet policing devoted to manage the allocation of packets eligible for the forwarding across a set of available queues (buffers), on the base of the priority assigned by marking function. Generic objectives of a scheduling algorithm include: fairness, i.e. the capability to minimize queuing time among traffics with equal profile; balancing, i.e. the capability to use the whole pool of resources available (optimizing, then both throughput and delay performances). Shaping: The function of modulating the service flow rate at the source, based on a feedback signal coming from the police system (classifier, meter, scheduler) receiving that service flow. Shaping function is an optional process used to avoid or minimize congestion event at a scheduling sub-system downstream. A.1.8 Subscriber data and session management Fixed network: No final report - Waiting for acceptance from the European Commission Page 101 of 119

102 Session (Broadband Forum Working Text WT-146): A Session is a logical construct intended to represent a network connectivity service instance at a network node. Data and control plane policies are associated with Sessions. Sessions are initiated and configured dynamically or statically. A Session may have associated state. Subscriber session (WT-146): A Subscriber Session is a PPP Session, an IP Session, or an Ethernet Session. Subscriber sessions are used to represent all traffic that is associated with that subscriber by a given service provider in order to provide a context for policy enforcement. Subscriber session creation (WT-146): Subscriber session creation is the mechanism to create a new subscriber session, by provisioning for static subscriber sessions or by detecting triggers for dynamic subscriber sessions. For example, the following triggers define a subscriber session start (non exhaustive list): an Ethernet 802.1ad or 802.1Q packet received by the Service Edge for an Ethernet session; a DHCP packet received by the IP Edge for an IP session. Subscriber session detection (BBF Technical Report TR-92; TR-101): The session detection is used for PPP session detection and allows a PPP session creation. Circuit ID and remote ID insertion (IETF RFC3046, BBF TR-101): The circuit ID and the Remote ID characterize the access loop and allow a service provider to identify a user and to apply specific service parameters. For a 1:1 deployment scenario, the Circuit ID and the Remote ID are inserted by the BNG, performing DHCP relay agent for IPoE access method (DHCP Option 82). For a N:1 deployment scenario, the access node inserts the access loop information: PPPoE intermediate agent for PPPoE access method, DHCP relay agent for IPoE access method. Session identification: The session is identified from the subscriber session identifiers derived directly from the packet that triggers the session creation. The session identifiers is a combination of (not limited to) Circuit ID and/or Remote ID, DHCPv4 option 60, source MAC address of DHCPDISCOVER message, packet s source IP address, packet s source MAC address, PPP identifiers No final report - Waiting for acceptance from the European Commission Page 102 of 119

103 Session Authentication: Session authentication is a mechanism to verify the identity of the user whom the session belongs to: for example, the authentication of a PPP session with CHAP protocol or the authentication of an Ethernet session with 802.1X. Credentials for session authentication can be a login/password or identifiers in a SIM card. Session Authorization: Session authorization performs a checking of the static rules defined by the network operator. If Resource Admission Control is not supported, session authorization allows traffic communication. (Step after session identification or session authentication). IP configuration: IP configuration relies on configuration of the network parameters of end devices or Residential gateway during the network attachment procedure: for example IPv4 address, IPv6 prefix, default route, DNS server DHCP, SLAAC, PPP/IPCP are such protocols able to configure network parameters. Session accounting: Session accounting is a mechanism which provides information on subscriber session traffic to the AAA server which counts the traffic during the session duration. Session accounting uses the start/interim/stop messages from the AAA client and is triggered by the session creation/monitoring/termination. Session monitoring: The session monitoring can provide an accurate picture of the state of a subscriber session. For Ethernet sessions, Ethernet OAM functions provide tools to check session continuity. IP sessions can be monitored thanks to L2 underlying connectivity or thanks to Bidirectional Forwarding Detection (BFD) echo keep-alive. Session modification: Session modification is the modification of one parameter of the session during the session duration, for example the IP address or the traffic policy. Session modification can be done through management system or through AAA after being triggered by a specific event (eg quotas). Session termination: No final report - Waiting for acceptance from the European Commission Page 103 of 119

104 A session can be terminated by management plane, control plane or data plane. For dynamic sessions, session termination is governed by the subscriber policy/termination rules, which can be a for example failure of a keep-alive protocol. AAA client: The AAA client communicates with the AAA server through RADIUS or DIAMETER protocol. It sends AAA Start/Interim/Stop messages for each session depending on the state of the session. It also receives the policy rules from the AAA server. Session database: Give a view of all sessions and their associated status. Mobile network: In mobile network, subscriber data management handles subscriptions related information, such as user identity and security credentials required for an end-user device to connect to a LTE/EPC network and performs authentication. Also, subscriptions information may differ in terms of allowed radio access technologies, services and QoS available, real time charging (pre-paid) payment or not (post paid), charging model for the data consumed, etc. Since in this study, the functions related to the subscriber session management are included in this functional group, these can be also described in the context of a device attach procedure. The Figure shows a simplified sequence of the attach procedure of a device making use of a mobile radio access according to the 3GPP E-UTRAN access [5]. Through this procedure, a mobile device can register its location in a serving MME. The network configures a signalling radio bearer, which carries the subsequent non access stratum signaling messages across the air interface. The network also provides an IPv4 / IPv6 address to the device and sets up a default EPS bearer, which provides the mobile with always-on connectivity to a default Packet Data Network (PDN). In the Figure, the dashed lines indicate functionalities that are optional or conditional. For more information please refer to [5]. No final report - Waiting for acceptance from the European Commission Page 104 of 119

105 Figure 31: Mobile Network: attach procedure for devices. Attach Request: The execution of this function aims to register a mobile device in a serving MME. It is triggered by a device s message, including a globally unique temporary identity (used when last switched on), the identity of the tracking area (last registered location), as well as information from the device's non access stratum capabilities and supporting security algorithms. Identification, Authentication and Security: These functionalities check the subscription data information associated to the device and also the validity of its request to the network. Location Update: The location update function permits the network to update and keep record of the devices' location. Session Creation: A session or an EPS bearer provides a logical transport channel between the UE and the PDN for transporting IP traffic. The session creation function is performed by a MME once it has all the information requires to set up the default EPS bearer. Attach Accept: This function completes the reply from the MME to the device's attach request. Here, examples of information provided are EPS bearer identity, device s access point name, quality of service and any IP address that the network has allocated to the mobile. No final report - Waiting for acceptance from the European Commission Page 105 of 119

106 Session Reconfiguration: The session reconfiguration functions in this context correspond to a Default Bearer Update. Once this function is performed, e.g. the downlink data packets can flow to the mobile. A.1.9 Mobility This group includes the mobility of the device/user and the nomadism (access to the network from different locations). Mobility is the ability to support from low to high speed users, providing a seamless service experience and session continuity. Specific functionalities included in this group are: Mobility Client: This function is for wireless mobile networks like those standardized by 3GPP, is limited to the specific case of interworking with non-3gpp networks like those standardized by IEEE. The most significant Mobility Client example in this case is related with the ANDSF node and communications protocol, where an ANDSF mobility client in the mobile terminal communicates with the ANDSF node in order to decide the best network to provide a service to the terminal, based also in service policies provided by the PCRF. Within the IEEE standards, and for providing nomadism capabilities to the terminal, it can include a client for simple Access Point discovery and attachment, being the solutions proposed by the Wi-Fi Alliance s HotSpot 2.0 a typical example [8][15][16]. Mobility Anchor: In the case of 3GPP networks, it usually refers to the Radio Access Network (RAN) node where the terminal is originally attached to when the mobility procedures are performed and who is in charge of transferring the terminal attachment to another node. One example of this mobility anchor node in 3G/UMTS is the Radio Network Controller (RNC). Mobility Decision: The implementation of this generic functionality varies widely depending on the type of access network. In the case of 3GPP wireless networks, like 3G or LTE, mobility decisions are implemented in the NodeB/eNodeB, based on radio status reports from the mobile terminal. Measurement and Reporting: No final report - Waiting for acceptance from the European Commission Page 106 of 119

107 Measurement and Reporting of Radio Link Quality is a functionality restricted to the physical layer of the mobile networks, where a set of protocols are used by the terminal to indicate some periodic or on demand measurements that are reported back to the base station. These measurements are later used to take mobility decisions at the base station. Paging: This functionality is restricted to mobile wireless networks, where there is a need to address a specific terminal whose precise location is only known to a limited extent. Paging is based on keeping coarse location registration of the terminal, by means of location or tracking areas, and sending paging messages to the terminal in that area. Localization: This functionality can range from broad range localization, based on tracking area localization or base station connection localization, to fine localization based on intra- 3GPP techniques or leveraged on other technologies like GPS or Wi-Fi s Access Point detection. A.1.10 Legal interception and data retention Legal interception: Communications network data of a specific fixed or mobile access connection have to be forwarded to a law enforcement agency for the purpose of analysis or evidence gathering. Legal interception typically requires the following functions: Access line identification (Identification of the access line of the end user to be intercepted); Content decryption (if it is encrypted); Content duplicating; Content encryption with a standard key (optionally); Content forwarding to law enforcement agency. Emergency calls: The caller location must be ascertainable even if the caller is unable to talk and therefore cannot give his name or location. The Emergency Call function enables the connection of emergency calls to pre-determined emergency units, e.g. police, fire department or medical personnel. Emergency call typically requires the following functions: Access line identification (Identification of the physical location of the calling end); No final report - Waiting for acceptance from the European Commission Page 107 of 119

108 Priority for emergency calls; Route emergency calls to recorded multilingual announcements. A.1.11 Traffic analysis The tools available for traffic monitoring can be considered as part of the Operation Administration and Maintenance (OAM) framework. The simplest tools available for traffic analysis rely on counters present in most forwarding equipment, that meter incoming and outgoing traffic flows. These counters can be regularly downloaded using OSS tools such as e.g. SNMP. More advanced tools are implemented by inserting extra packets in the data flows and analyzing the contents of these measurement packets. Others rely on inserting test traffic in order to probe network performance. A last category of tools relies on sniffing actual traffic (e.g. copying either packet headers or complete packets) at specific interfaces. Those tools include: Inserting, extracting and analysing in band OAM flows (e.g. IEEE 802.1ag and ITU-T Y.1731 for Ethernet OAM, IETF RFC 6371 for MPLS-TP); Implementing out of band OAM flows (e.g. PING messages in IP networks); Inserting dedicated test flows thanks to a dedicated measurement equipment; Performing packet header analysis, either on all live packets, or on a given proportion of live packets (e.g. NetFlow); Performing Deep Packet Inspection (DPI) in order to analyse the complete contents of the live packets. A.2. Detailed description of equipment As part of a preliminary functional analysis of current networks, COMBO WP3 has listed, for all equipment considered in section 2.1, the main functional groups and functions implemented in the equipment. This detailed description of equipment was based on the functional groups and functions described in appendix A.1. This work was collected in several sheets of a spreadsheet file, and was too heavy to be directly inserted as tables in this D3.1 document. The interested reader can thus directly refer to this spreadsheet file for a detailed functional description of equipment in current networks [65]. A.3. Overview on Mobile Network Architecture The diagram depicts the architecture of current 2G, 3G and 4G/LTE mobile communication networks. Devices in the communication network are called nodes. No final report - Waiting for acceptance from the European Commission Page 108 of 119

109 Nodes are connected by interfaces (also called reference points). Mobile devices are using radio interfaces. Nodes within the fixed network communicate via interfaces based on copper or optical cable as physical medium. Figure 32: Overall architecture of mobile communication networks In 2G a mobile device is called a Mobile Station (MS). Mobile stations connect to a Base Transceiver Station (BTS) via the Um radio interface. BTSs connect to a Base Station Controller (BSC) using the Abis interface. BTSs and BSCs comprise the GERAN (GSM / EDGE Radio Access Network). No final report - Waiting for acceptance from the European Commission Page 109 of 119