A Framework to Improve QoS and Mobility Management for Multimedia Applications in the IMS

A Framework to Improve QoS and Mobility Management for Multimedia Applications in the IMS Fabricio Carvalho de Gouveia Technical University of Berlin Franklinstr. 28-29, D-10587 Faculty IV / Next Generation Networks Berlin, Germany gouveia@cs.tu-berlin.de Thomas Magedanz Technical University of Berlin Franklinstr. 28-29, D-10587 Faculty IV / Next Generation Networks Berlin, Germany tm@cs.tu-berlin.de Abstract The new IP Multimedia Subsystem (IMS) provides an overlay architecture for IP based core networks and enables the efficient provision of an open set of potentially highly integrated multimedia services (I.e. instant messaging, video conferencing, VoIP, application sharing, etc) on top of possibly different access technologies. However, there will be different domains and IMSs managed by different ISPs (Internet Service Providers) in the Internet. Each ISP has an SLA (Service Level Agreement) established between its users, which defines what kind of resources and prices was agreed to be offered to each user. However there is no SLA between a visited domain and a mobile user. This paper analyses how to extend the QoS capabilities of the IMS offered to a mobile terminal when it roams among different domains. We propose a framework that integrates QoS and Mobility management. This integration envisages the improvement of the overall end-to-end performance and the reduction of management entities in the network. The architecture presents the Domain Policy Manager (DPM) as the Decision-making and controller of a domain. The proposed architecture uses policies with inter-domain negotiation. An application scenario to show the framework functionality is described as well as the current Status of the implementation. 1. Introduction The Internet is composed by different domains, each of them managed by different ISPs (Internet Service Providers). These Domains have different access networks and capabilities, users and policies that rule their behavior. Each ISP has an SLA (Service Level Agreement) established between its users, which defines what kind of resources and prices was agreed to be offered to each user. However there is no SLA between a visited domain and a mobile user. Thus there is a need to address inter-domain QoS negotiation to ensure that sufficient resources are provided to mobile users. With the new emerging area of IP Multimedia Subsystem (IMS) [1] and Next Generation Networks (NGN), QoS and mobility support are essential for the success of such networks. The concept of NGNs is based on the convergence of fixed and mobile telecommunication networks and the Internet towards an all-ip environment. These NGNs support the provision of integrated information and communication services in face of increasingly more complex value chains, enabled by so-called service delivery platforms (SDPs). Therefore an NGN is an environment of high complexity, in which different actors, such as fixed and mobile network operators, service providers, system integrators and an open set of application providers have to cooperate for the provision of advanced converged services. A converged service is the integration of voice, multimedia, web content and Web Services provided seamlessly over all kinds of access technologies. This heterogeneity increases the complexity and challenges related to the management of such networks. The specific problem addressed by this research is how to ensure necessary levels of service and performance to critical multimedia and data applications, including service guarantees, while automating management. This scenario is illustrated with an inter-domain vertical handover, where the QoS must be guaranteed with an acceptable level while a mobile user roams through domains. Our main contribution is the proposal of a framework that addresses end-to-end QoS in the case of user mobility. We propose to integrate the management of QoS and mobility at the IP level. For

such mechanisms to be efficient, we need an efficient mobility management scheme optimized for QoS. Integration of mobility management with QoS can bring advantages to the system, for instance, it is possible to take into account a richer set of parameters to initiate handovers. The decision to switch to another cell can be made not only based on the signal to noise ratio, but also on the current load in a cell, the level of available resources and the state of pre-reservations or administrative policies. We claim that this integration can improve the overall end-to-end performance and the reduction of management entities in the network. The content of the paper is structured as follows: First we discuss the policy architecture in the IMS, which we will enhance with the concepts of our framework and the details of FOKUS 3Gb Testbed [2], a platform for testing and validating the underlying framework. Section 3 shows the requirements that NGNs impose on the management plane. Related work is commented in section 4. In section 5 we present the conceptual model of the QoS and mobility framework, its components and main capabilities. An application scenario is described in section 6 and discussions about the implementation in section 7. Finally, the conclusions and future work are presented in section 8. 2.1 Policy Architecture in the IMS 3GPP adopted a policy-based QoS solution to ensure that sufficient QoS resources are provided to authorized users. The reference model of a policybased network consists of two main elements, the Policy Decision Point (PDP) and the Policy Enforcement Point (PEP) [4]. The PDP weighs the policy request sent by the PEP, as a result of a policy event against a corresponding set of policy rules. As a response to a policy request, the PDP either evaluates the policy rules for the request, which is referred to outsourced policy or retrieves the set of policy rules relevant for the request, which is referred to provisioned policy (The Provisioning and Outsourcing policy models are described in [5]). The policy decision or the set of policy rules is then transported to a PEP using the policy transaction protocol Common Open Policy Service (COPS). The PDP is the final authority the PEP needs to refer for actions to be taken. 2. The IP Multimedia Subsystem The IP Multimedia Subsystem (IMS) [1] as part of the 3GPP Release 5 specifications defines an overlay architecture on top of the 3GPP Packet Switched (PS) Core Network for the provision of real time multimedia services. The IMS is based on principles and protocols defined for the Internet by the IETF (Internet Engineering Task Force), which have been adapted for their use within a secure, scalable carrier grade environment. The Session Initiation Protocol (SIP) is used as the signalling protocol that establishes, controls, modifies and terminates voice, video and messaging sessions between two or more participants. In the context of the IMS Architecture the related signalling points are referred to as Call State Control Functions (CSCFs) and distinguished by their specific functionalities. Functionalities related to Authentication, Authorization and Accounting (AAA) within the IMS are based on the IETF Diameter [3] protocol and implemented in the HSS, CSCFs and various other IMS components in order to provide charging functionality. The HSS forms the organizational glue between the CSCFs and the Application Servers (ASs), whereas the real time signalling between CSCF and ASs is based on SIP. Figure 1- Policy Architecture in the IMS The PDF can be a logical component of a Proxy- CSCF (P-CSCF) (Release 5) or a separate entity altogether (Release 6). Since the GSM Gateway Support Node (GGSN) is in the data path, it is the logical location for the PEP. The policy repository can be an entity external to the PDF. A Lightweight Directory Access Protocol (LDAP) capable data store can be used for this purpose. The PDF communicates with the PEP via the Go interface [6]. Figure 1 depicts the relationship between these entities. 2.1.1 Proxy-CSCF. The Proxy-Call Session Control Function (P-CSCF) is the first contact point for users within the IMS. All SIP signalling traffic from or to the UE go via the P-CSCF. The P-CSCF validates the request, forwards it to selected destinations and processes and forwards the response. There can be one or many P-CSCFs within an operator s network. Please refer to [7, 8] for the detailed descriptions of the functions performed by the P-CSCF.

2.1.2 PEP in the GGSN. The role of the PEP is to ensure that only authorized IP flows are allowed to use network resources that have been reserved and allocated to them. The PEP in the GGSN is responsible to drop the IP flow that was not permitted by the PDF. This process is called policy-based admission control. This process ensures that an IP flow is only allowed to use resources that have been approved by the policy rules. The PEP may store decisions in a local PDP. In this case, the GGSN can make admission control decisions without additional interactions with the PDF, reducing this way, the traffic over the Go interface. networks, such as GSM, WLAN, digital video broadcasting, UMTS, etc. including the related end systems. Next to the network layer is the control and management layer, constituted by signaling and management components. This layer is responsible to provide the IMS capabilities like QoS, security, AAA, etc. forming the service delivery platform. The highest layer is the application layer, where the services are developed. 2.1.3 User Equipment (UE). The UE obtains an authorization token from the P-CSCF via SIP signaling during session setup. This token is used to provide the binding mechanism that associates the PDP context bearer to the IP flow in order to support IP policy enforcement in the GGSN. By examining this token received from the GGSN, the PDF can direct the GGSN to admit or drop the flow. 2.1.4 The Go and Gq Interfaces. The Go interface allows service-based local policy and QoS interworking information to be requested by the GGSN from a Policy Decision Function. This allows operators to control QoS in a user plane and exchange charging correlation information between IMS and GPRS network. The protocol used for this is the Common Open Policy Service (COPS) [5]. The COPS protocol is a simple query and response protocol that allows policy servers (PDPs) to communicate policy decisions to network devices (PEPs). The protocol uses Transfer Control Protocol (TCP) to provide reliable exchange of messages. COPS provides the means to establish and maintain a dialogue between the client and the server and to identify the requests. The Go interface provides information to support the following functions in the GGSN [6]: Control of service-based policy "gating" function in GGSN; UMTS bearer authorization; Charging correlation related function. The Gq reference point is used to exchange policy decisions-related information between P-CSCF and PDF and is being standardized in 3GPP Release 6 [1]. The protocol specified for this interface is Diameter. 2.2 The IMS Playground at FOKUS Testbed FOKUS 3Gb Testbed consists on three logical layers as shown in Figure 2. The lowest layer is the network plane that integrates wireless and fixed access Fig. 2 Open IMS within the 3Gb FOKUS Testbed. FOKUS researches and develops the building blocks needed both for end-to-end seamless integration of technologies and end devices for the deployment of open flexible communication services and applications. The core elements of the 3G beyond Testbed comprise of playgrounds for Multimedia Value-added Service Platforms covering the whole range of relevant telecommunication service platforms for converging networks, like OSA/Parlay and IP Multimedia Systems. The Open IMS provides a platform for the validation of our QoS and mobility framework. 3 Requirements for the Management of Next Generation Network As discussed in the introduction, NGNs puts new burdens in the management of these kinds of heterogeneous networks (different wired and wireless technologies). This section summarizes the requirements for the management of QoS and Mobility: QoS Requirements:

To guarantee QoS among different domains and networks; End user perception of QoS. Mobility Requirements: ability to change access point and/or terminal; ability to get access from any network access point, including all access technologies identified; ability to get services in a consistent manner, subject to the constraints experienced in their current situations; The framework must be aware of user availability. the mobile node to perform handover only if enough resources are available. 5 Conceptual Model of the QoS and Mobility Management Framework The reference model of the framework consists of two main elements, the Domain Policy Manager (DPM), which is an extended version of the PDF, acting as a Policy Decision Point (PDP) and the Policy Enforcement Point (PEP). This model is depicted in Figure 3. The components of this conceptual model structure are based in the Internet Engineering Task Force (IETF) policy framework group [16]. 4 Related Work In Recent years Policy-based management has become a great topic of research. However, most research in this area focus in intra-domain QoS or Security. There is no much work with Policy Based Management (PBM) that addresses mobility issues. A Study in 3GPP [9] address the WLAN-3GPP system interworking and specifies 6 scenarios with different levels of integration. Policies are used to control the bearer traffic on 3GPP IMS. PBM addressing mobility hasn t been the subject of as many works as the security or QoS ones. It is difficult to meet all the requirements for automated management in a heterogeneous system. However, there are some works trying to define a model to use policies in the mobility management. In [10] the authors propose extensions to the COPS protocol, called COPS-MU (Mobile User) and COPS-MT (Mobile Terminal), to deal with terminal and user mobility. They define new policy objects in the COPS protocol for terminal and user registration. There is another extension [11] called COPS-SLS (Service Level Specification) to support the QoS negotiation. We claim that an integrated Policy Architecture can support mobility and QoS management if there is cooperation between DPMs. The DPM can be compared to the Bandwidth Broker (BB) in the Differentiated Services (Diffserv) Architecture [12]. The European research projects AQUILA [13] and TEQUILA [14] use a centralized QoS approach like the DPM. With a single control point, the integration with different mobility schemes and location management becomes more flexible [15] The DPM is a centralized manager that handles the resources of one domain. It keeps the actual stand of resources and reservations in its domain. The authors motivate to use a central approach in order to support anticipated handover with pre-reservations, allowing Fig. 3 Conceptual Model of the QoS and mobility framework. The role of the PEP is to ensure that only authorized IP flows are allowed to use network resources that have been reserved and allocated to them. The PEP in the Access Router (AR) is responsible to drop the IP flow that was not permitted by the DPM. This process is called policy-based admission control. This process ensures that an IP flow is only allowed to use resources that have been approved by the policy rules. The PEP may store decisions in a local PDP. In this case, the AR can make admission control decisions without additional interactions with the DPM, reducing this way, the traffic between them and lessening the processing load on the DPM. The modules of the DPM are: DPM Management Tool: Is the interface for the domain administrator to add, edit and remove policies, as well as to control the

Policy Repository for consistency and search for policies conflicts. Policy Repository: Is the entity where all the policies of a domain are stored (for home users, as well as policies that rule what to do when a foreign user joins the domain). Resource Negotiation Function: This is the most important function of the DPM. It is responsible to communicate with another DPMs to reach an agreement for requests that the domain can t provide for a roaming user. It searches for common parameters between the domain and the user. Resource Monitoring Function: Monitors the network for the current status of resources and availability of users. It ensures that the contracted services are being met. 5.1 Design Considerations 5.1.1. Interoperability with other architectures. In an end-to end scenario there are several administrative domains, each with its own policies, resource management architectures and traffic mechanisms. A common protocol would be needed for communicating end to end the QoS requirements of user traffic, while at the same time respecting the individualities of the autonomous operation of each traversed domain [17]. However it is not possible with different autonomous networks and many proprietary solutions. To find a solution to interoperation of the system, it is indispensable to separate the signaling protocol from the carried in-formation. In this way, control signaling can he carried by any signaling protocol, while being understood by any autonomous system. The concept of our framework follows this standard, decoupling the network control plane from the packet forwarding plane. 5.1.2. Signalling Protocol. An inter-domain signalling protocol is required in order to ensure interaction among the managers, in our case the DPMs. The protocol could be an integrated protocol in order to manage and coordinate the QoS and mobility functionalities. We searched solutions in the literature to facilitate the management activities. An example for QoS and routing functionalities can be found in [21], but a more complete solution [22] proposes to extend COPS to provide the interaction among the dedicated management protocols (mobility, QoS and security). We choose to use this COPS extension solution for inter-domain negotiation in our framework. 5.1.3. Integration Level. There are two level of integration in the literature: tightly coupled architecture and loosely coupled architecture. There are advantages and disadvantages when the two architectures are compared. The main difference between these two approaches is the point where the networks are connected in the network architecture. In a tightly coupled architecture, the WLAN network is connected to the UMTS network as an alternative radio access network. The data sent by user devices must go through the UMTS PS (Packet Switched) domain served by the connecting SGSN (Serving Gateway Support Node) to reach its destination. The session control entities like the CSCFs interact directly with the user devices as if they are normal UMTS user equipment (UE). This way the DPM can enforce the network-level policies at the PEPs directly as if they were part of the UMTS PS domain. There is no effect on the 3GPP access control and billing/charging entities. In a loosely coupled architecture, the user devices are is connected to a GGSN (GSM Gateway Support None), and are considered as peers of the UMTS network. A tight integration may result in a minimum data exchange due to the reduction of decision messages among other components. We decided to select the tight integration for our framework. The main factors are the total system delay for QoS, which must be the minimum possible since we are considering the management of multimedia applications and the handover that will be faster with less exchange of messages among entities. 6 Application Scenario In this section, we give an example scenario of an inter-domain vertical handover (see Figure 4). In this scenario there are two IMS domains: Domain A, which is the home network of the user and the visited Domain B, in which the user will perform the movement. The User s Equipment (UE) is a Personal Digital Assistant (PDA) attached to a dual access card (WLAN - GPRS) and therefore it can support this type of handover. The SIP signaling and registration process in the IMS is not demonstrated here, but is described in [1]. The focus is on the inter-domain QoS and mobility signaling.

Fig. 4 Inter-domain Vertical Handover. As the user is in his home domain, he has a Service Level Agreement (SLA) with his provider and requests the contracted resources, for instance, he wants to make a VoIP call. The DPM process the call admission control, reserves the required resources and authorizes the session in the local domain. Then, he starts the movement to the Domain B. The DPM in the home domain sends a service request message to the other DPM with the format of the QoS request, such as the type of service (VoIP) and QoS requirements that were authorized in the home services. The DPM on Domain B receives this request and access the Resource Monitoring Function, in order to know if there are sufficient resources to reserve for the user. If not, it informs the previous DPM about this state. If the resources are enough, the Visited DPM will check its Policy Repository to check the rules that will apply in this domain. If there are parameters that must be changed in Domain 2 (The technologies are different), the Resource Negotiation Function will start the negotiation with the DPM of Domain A. If an agreement is met, the DPM of Domain B will make the local decision based on the incoming service request message. If not the procedure fails and the session is finished. The PEP in Domain B will enforce these policies and the handover will be performed. Immediately the PDP of Domain B will send a Report State message back to Domain A. As the DPM receives this message, the whole procedure finishes. 7 Implementation The actual realization of an IMS core network proved to be a difficult task. First performance results can be founded in [18]. We are following specifications of 3GPP release 6 for all the implementations. The CSCFs are built upon the SIP Express Router (SER) [19] which can act as SIP registrar, proxy or redirect server and is capable of handling many thousands of calls per second. It is an open source SIP server implemented in the programming language C and has a modular structure that permits to make functionality additions. It successfully runs on many Unix-based platforms like Linux, BSD or Solaris. The P-CSCF is the first contact point to the DPM (through Gq Interface). In the current implementation of the P-CSCF, it is able to firewall the core network at the application level: only registered endpoints are allowed to insert messages inside the IMS network and the P-CSCF asserts the identity of the users. To keep track of the registered users, it has an internal registrar that is updated by intercepting the registration process and later by subscribing to the registration package at the S-CSCF and receiving notifications. The actual data is kept as a hash-table to allow fast retrieval. For originating call signaling it generates unique charging vectors and inserts network and path identifiers that are needed for the correct further processing of the SIP messages. UE forged information that might lead to an attack is removed. After a successful registration process to an IMS home network, subsequent user messages are forwarded based on DNS information towards the requested IMS network that will request the authorization token from the DPM. The DPM is being implemented in C Language. The Go interface is based on a client and server open source von VOVIDA [20] that is being extended to be implemented in the Testbed. The Gq interface is not implemented yet. 8 Concluding Remarks and Future Work Along this paper we presented the QoS and mobility framework for the management IP Multimedia Subsystems. The management of QoS and mobility is made at the IP level. The management of NGNs needs abstraction to handle heterogeneity, and we claim that policy-based management can bring the sufficient level of abstraction and automation needed in this field. The framework approach uses a centralized manager (one Policy Domain Manager per domain), which makes this framework easier to integrate with different mobility schemes in the literature. Another feature is the decoupling of the control plane from the packet forwarding plane. This allows interoperability with other autonomous systems because the information control can be carried by any signaling protocol. We demonstrated the functionality of the proposed architecture in an inter-domain vertical

handover scenario and described the current status of the implementation. With this integrated framework we can have a richer set of parameters to initiate handovers, which introduces new capabilities to the IMS. This way, the handover is only performed after negotiation of the resources. Acknowledgement I would like to thanks all members of Next Generation Networks and Infrastructures (NGNI) Competence centre at Fraunhofer Institute for Open Communication Systems (FOKUS) for all kind of support and help that I received during this research. References [1] 3GPP, TS 23.228. IP Multimedia Subsystem; (stage 2), may 2005. www.3gpp.org [2] www.fokus.fraunhofer.de/national_host [3] IETF RFC 3588, Diameter Base Protocol, September 2003. [4] IETF RFC 2753, A Framework for Policy-based Admission Control, January 2000. [5] IETF RFC 2748, The COPS (Common Open Policy Service) Protocol, January 2000. [6] 3GPP TS 29.207 V6.1.0, Policy control over Go interface, September 2004. [7] 3GPP TS 24.229 V6.4.0, Internet Protocol (IP) multimedia call control protocol based on Session Initiation Protocol (SIP) and Session Description Protocol (SDP); Stage 3, September 2004. [8] M. Poikselka, G. Mayer, H. Khartabil, A. Niemi, The IMS: IP Multimedia Concepts and Services in the Mobile Domain, Wiley, April 2004. [9] 3GPP, TS 22.934, Requirements on 3GPP system to Wireless Local Area Network (WLAN) interworking, September 2004. www.3gpp.org [10] G. Pujolle, and H. Chaouchi. A New Policy Based Management of Mobile Users, Net-working 2002, LNCS 2345, pp. 1099-1104, 2002. [11] G. Pujolle, and H. Chaouchi. QoS, Security, and Mobility Management for Fixed and Wire-less Networks under Policy-Based Techniques. IFIP World Computer Congress, 17th edition, August 25-30, 2002, Montréal, Canada. [12] S. Giordano, S. Salsano, S. V. Bergue, G. Ventre, D. Giannakopoulos, Advanced QoS Provisioning in IP Networks: the European Premium IP Projects. IEEE Com. Mag., Janu-ary 2003. [13] T. Engel et al., AQUILA: Adaptive Resource Control for QoS Using an IP-Based Layered Architecture, IEEE Communications Magazine, Jan. 2003. [14] E. Mykoniati et al., Admission Control for Providing QoS in DiffServ IP Networks: the TEQUILA Approach, IEEE Communications Magazine, Jan. 2003. [15] N. S. Smith, A Distributed Policy-based Network Management (PBNM) system for Enriched Experience Networks (EENs), Doctoral Assessment, November 2003. [16] draft-ietf-policy-framework-00.txt, M. Stevens, W. Weiss, H. Mahon, B. Moore, J. Strassner, G. Waters, A. Westerinen, and J. Wheeler, "Policy Framework," Policy Framework, 13 Sept 99. [17] S. I. Maniatis, E. G. Nikolouzou, L. S. Venieris, End-to-End QoS Specification Issues in the Converged All-IP Wired and Wireless Environment, IEEE Communications Magazine, June 2004. [18] D. Vingarzan, P. Weik, T. Magedanz, Design and Implementation of an Open IMS Core,Mobility Aware Technologies and Applications, MATA 2005, To appear on the proceedings by 16th October. [19] The SIP Express Router - http://www.iptel.org/ser [20]COPS Client from VOVIDA http://www.vovida.org [21] S. Lee et al. INSIGNIA: In-band signaling support for QoS in Mobile ad-hoc Networks. Workshop on Mobile Multimedia Communication, 1998. [22] H. Chaouchi et al, A Trial towards Unifying Control Protocols: COPS versus Radius/Diameter and Mobile IP MWCN 2002.