Mobility Management Reconsideration: Hierarchical Model and Flow Control Methodology

Transcription

1 Mobility Management Reconsideration: Hierarchical Model and Flow Control Methodology Jun-Zhao Sun and Jaakko Sauvola MediaTeam, Machine Vision and Media Processing Unit, Infotech Oulu P.O.Box SOINFO, FIN University of Oulu, Finland Abstract Mobility management is the fundamental technology to enable the seamless access to wireless networks and mobile services. Mobile IP is currently the most promising solution for mobility management in the Internet. The objective of this paper is to reconsider Mobile IP based mobility management solutions, with the emphasis on the proposal of new mobility management architectures. First, it is recognized that Mobile IP lacks flexibility with respect to the transparent and mandatory mobility support to various applications. This leads to a flexible hierarchical architecture proposal on the basis of optional mobility mechanisms at different protocol layers. Another problem occurs when multiple interfaces are available to one user device to access heterogeneous networks. In this case flow level traffic control is needed in addition to device level mobility management. A novel architecture for the multi-connectivity management is proposed to address this problem. I. INTRODUCTION Mobility management is the fundamental technology for the support of seamless accessing to wireless networks and mobile services. Two main aspects need to be considered in mobility management, i.e. location management (e.g. addressing, location registration and update, tracking and paging, etc.) and handoff management (e.g. handoff initialization, resource relocation, connection re-establishing, etc.). Future mobile communication systems evolve with the trend of global connectivity through the internetworking and interoperability of heterogeneous wireless networks. Roaming in such network systems is a more complicated situation that causes many new problems. The requirement of smooth and adaptive delivery of real-time and multimedia applications makes the design of mobility management scheme more severe a challenge that needs to be carefully considered. IP plays a crucial role for mobility management in various types of wireless networks as the access systems to the Internet. Great efforts for protocol standardization have been made in IP-based mobile telecommunications networks. In the movement toward 3G wireless, the two partnership projects that address the issue of standard development, 3GPP and 3GPP2, are both moving toward an all-ip mobile network architecture. IP is positioned on network layer, which makes it an isolator to prevent upper layer protocols and applications from the awareness of location changes. So IP is regarded as a suitable layer to solve the problem of mobile routing and provide transparent mobility support to applications and higher level protocols like TCP. The IP extensions for solving mobility issues are mainly carried out at the working group of IP Routing for Wireless/Mobile s (mobileip) in the Internet Engineering Task Force (IETF). The main goal of the mobileip working group is to develop routing support to permit IP nodes (hosts and routers) using either IPv4 or IPv6 to seamlessly roam among IP subnetworks and media types. The proposed Mobile IP method supports transparency above the IP layer, including the maintenance of active TCP connections and UDP port bindings. A large number of standard proposals, in form of both Internet Draft and Request For Comments (RFC), have contributed to different aspects of mobility management issues. Mobile IP is currently the most promising solution for mobility management in mobile Internet. The objective of this paper is to reconsider Mobile IP based mobility management technology, with the emphasis on the proposal of new mobility management architectures. First, it is recognized that Mobile IP lacks of flexibility with respect to the transparent and mandatory mobility support to various applications. This leads to a hierarchical architecture proposal on the basis of optional mobility mechanisms at different protocol layers. Another problem occurs when multiple interfaces are available to user device for the accessing of heterogeneous multi-access networks. In this case flow level traffic control is needed in addition to device level mobility management. A novel architecture for the multi-connectivity management in the multi-access and multihoming context is proposed to address this problem. The rest of this paper is organized as follows. Section II presents the hierarchical mobility management model, based on the analyses of related research and new requirements. In section III the benefit, related research, and solution of flow level traffic control are discussed in detail. Section IV concludes the paper. II. NEW MOBILITY MANAGEMENT MODEL A. Related Research Mobility management can be implemented at different layers. Link layer solutions, e.g. GPRS and IEEE , are protocol-embedded solutions with lower overhead. They are suitable for movement within an administration domain but not feasible for larger movement. Network layer solutions, e.g.

2 Mobile IP [1], Mobile IPv6 [2] and its extensions [3-7] make mobility transparent to all upper layers including transport layers and applications. Lots of research has been carried out on e.g. routing optimization [3], fast and seamless handoff [4], hierarchical architecture [5-6], micro-mobility for intra-domain [7], etc. Transport layer solutions, e.g. TCP and DNS extensions [8], are end-to-end approaches without network layer support. Application layer solutions, e.g. SIP with mobility support [9], are used to support real-time multimedia applications. Moreover, mobility solutions at application level can also address the problem of personal and service mobility. The basic idea for a mobility management solution is shown in Figure 1. As illustrated in Figure 1 (a), when an end host moves into a new region and tries to connect to a new access point (step 2), it needs to first reconfigure itself to get a new exercisable address by address allocation mechanism. This configuration can be done either manually or automatically by e.g. Dynamic Configuration Protocol (DHCP) or IPv6 autoconfiguration process. A user may also need to do authentication at the new access network. After a topologylogical address is correctly obtained, the end host can then register or update the corresponding entry in the name-toaddress translation (N2AT) component (step 3), to make it reachable by other communicating peer hosts. The peer host can then obtain the new updated address (step 4) and form the new connection to the new access point (step 5). N2AT component realizes the most important function in the solution architecture. The N2AT database can either operate at independent equipment, or reside in the peer host. In practice, this function is mapping to Home Location Register (HLR) in cellular communication systems like GSM and GPRS, Home Agent (HA) in Mobile IP, DNS for transport layer solutions, and SIP redirection server for application layer solutions. Figure 1 (b) shows the effect of the mobility management implementation. A totally transparent mobility support is provided to upper protocol stack and applications. Mobility is modeled by references as changing node s point of attachment to the network, i.e. an address translation problem. So it is stated to be best solved at the network layer by changing the routing of datagrams destined to the mobile node to arrive at the new point of attachment [10]. To implement mobility management at network layer may also shield the upper-level protocols from the nature of the physical medium and make mobility transparent to applications and higher-level protocols. Currently network layer mobility management is mainly addressed in two different communities: the PCS community and the Internet community. The works in the PCS community focus on the effort on location and handoff management of a cellular phone [11]. Also many works have been done in the field of wireless ATM [12]. The works in the Internet community focus mostly on the standardization of Mobile IP aiming towards extending IP with the capabilities of dealing with mobility. Mobile IP has attracted mass attention due to the all-ip architecture of the new mobile systems. B. New Requirements Besides the basic functionality requirements on mobility management schemes, new requirements on mobility management technology are the following. End 2 End 3 N2AT 5 (a) General implementation model Peer Mobility-transparent application systems Mobility implementation Mobility-aware basic systems (b) Transparent mobility support Figure 1. Mobility management solution Application characteristics should be carefully taken into account during both the design and the utilization of a certain mobility support mechanism. Since the generality of mobility support is comes at significant cost, complexity, and performance degradation, it should remain an option even if it is implemented. It should be possible to decide whether an optional mobility support mechanism should be employed or not, and if it is, to what extent and in what form it will be utilized. This can be finally decided at run-time, automatically by the system or explicitly by the applications. As opposite to the manner of Mobile IP that tries to hide mobility from the transport layer and applications, there should be interfaces where applications may become aware of mobility if they want. Actually mobility-aware applications are becoming more and more popular in the new ubiquitous and pervasive computing paradigm. The deployment of mobility management mechanisms should be easy and smooth, which means to minimize the changes on current infrastructure. This leads to the question whether the mechanism should be as a protocol extension or a software packet based on standard and existing protocol technology. New architectures and solutions for mobility are needed to fulfill these new requirements, other than the fully transparent and mandatory solution based on Mobile IP. C. Hierarchical and Adaptive Model One serious limitation of previous Mobile IP based solutions is the inflexibility, as illustrated in Figure 1 (b). An application can be either mobility-aware or mobilitytransparent. A mobility-transparent application cannot perceive any connection changes happening, and so there must be some 1 4

3 mechanisms in the transport system to cope with the changing network context and keep the application unaware. On the contrary, a mobility-aware application is aware of the special events happening on network connections. Such events may include e.g. the plug-in or pull-out of a network interface card, varying availability of connectivity, new attachment point due to mobility, network QoS status changes, etc. These applications do not need a seamless and transparent management on mobility. Instead, they may hope to deal with the changes of location and connection by themselves. In this case the mobility management mechanisms should be somehow bypassed. We propose a new architecture model for mobility management, as shown in Figure 2. Compared to the general idea shown in Figure 1 (b), there are some new features and advantages of this hierarchical and adaptive model. The model is generic and needs to be specialized. A typical specialization of the model is that mobile-aware basic systems are the physical and link layers, the mobility implementation layers are, from bottom up, network layer solutions, e.g. Mobile IP and micromobility IP, transport layer solutions, e.g. mobile TCP or UDP, and application layer, e.g. mobile SIP. Mobility-transparent applications do not mean they are totally unaware of mobility. In pervasive applications it is important to make applications aware of and adapt to the changing network conditions due to e.g. mobility. Multiple solutions at different protocol layer can be coexistent in one communication system, and take into effect by different requirements of various applications. Cooperation between mobility mechanisms at different layers is also possible. In each layer, normal protocol implementation and mobility implementation are optional either by the applications themselves or automatically by additional methods. The new model emphasizes more higher level solutions at e.g. transport and application layers. Higher layer mobility solutions have many benefits compared to Mobile IP based network layer solutions, including More efficient. There is no need for HA or FA to forward datagrams. Data packets are routed normally end-to-end, so no abnormal routing happens. Easy to be deployed. Since higher level solutions do not require any changes to the IP stack and OS of any end host, they can be successfully deployed widely much easier than Mobile IP. Flexible. Higher level solutions mean that the implementations are totally end-to-end. It is flexible for applications to control the mobility mode according to their specific requirements. Enabling mobility-aware applications. This can be achieved through providing some new application programming interfaces (APIs), and thus enabling applications to learn about mobility and adapt to it accordingly. Mobility Transparent Mobility Mobility-aware basic systems There have been some higher level solutions for mobility management other than Mobile IP based network layer methods. Transport layer solutions, e.g. TCP and DNS extensions [8], are end-to-end approaches without network layer support. From an application s point of view, transport layer solution means that the opened sockets remain the same during the host moving. There must be a mechanism to trace the changing network IP addresses of the peer, e.g. with enhanced DNS operations. There are also some proprietary methods [13] that implement the seamless handoff by employing a proxy instead of by extending any existing protocol. This method may have the problem of breaking the end-to-end transport semantic. Application layer solutions, e.g. SIP with mobility support [9], are used to support real-time multimedia applications. Moreover, mobility solution at application level can also address the problem of personal and service mobility. III. Aware Mobility Application FLOW LEVEL TRAFFIC CONTROL Implementation Infrastructure Figure 2. Hierarchical mobility management model A. Benefits Future communication systems will be based on an all-ip core backbone infrastructure with heterogeneous access networks in order to realize global accessibility [14]. Diverse networking resources are converging together to fulfill user and application requirements. In Mobile IP the solution to this multi-access case is vertical handoff, i.e. handoff between different accesses. This results in a device level flow control where only one network interface can be active at a time. In the case when multiple interfaces can only be used exclusively, vertical handoff provides an effective mechanism to seamlessly switch the network connection when necessary. However, generally the case is that multi-access network connectivity can be available simultaneously, for example when a laptop equipped with both GPRS and WLAN wireless adapters enters into the overlapping areas of the two access networks. A straightforward question against the vertical handoff scenario is then, since multiple interfaces are simultaneously available, is it possible to activate multiple interfaces at an instant and use a set of them simultaneously? Benefits of the simultaneous usage of a combination of network connections include: Some services are connectivity-dependent, which means that a specific type of connection must be available. So one interface should not be masked by another one due to the presence of a mobility management mechanism as in Mobile IP.

4 The system can deliver each service via the access network that is most efficient and suitable for it, to provide a wide range of QoS to user. One application can have several connections and each traffic flow can select to use the access type that is best suitable to its characteristics. Each data flow can be dynamically allocated to and redirected between different interfaces to adapt to the dynamically changing network status and balance the workload. Access networks may be combined together in use to increase capacity in terms of network bandwidth and reliability, and decrease handoff delay. Different access networks might be used for asymmetric traffic, e.g. separate uplink and downlink. In case of unidirectional connectivity as in a satellite system, it s important to provide connectivity in reverse direction via a different access e.g. cellular. Multiple interfaces can be used simultaneously, complementarily, and collaboratively. Connectivities can be interoperable with each other to provide new styles of networking support, to enable novel pervasive applications. It is worth noticing that for a multihomed host the concept of mobility is extended to a higher level where even a fixed machine can also be mobile. This happens when a fixed host is equipped with multiple interfaces and dynamically redirects traffic flow from one to another. For example a desktop PC with both Ethernet LAN card and a dialup modem may switch data packets between the two interfaces. B. Related Research As we have explained, current solutions for this situation focus mainly on network layer, by using e.g. Mobile IP to implement upward or downward vertical handoff between heterogeneous access networks [15]. There are also some proprietary methods [13] that implement seamless handoff by employing a proxy. This method may have the problem of breaking the end-to-end transport semantics. There are some network layer solutions [16-19] concentrating on the Mobile IP BU extension to support multiple interfaces and data flow control. The extended BU can be sent to the MN s HA, GFA and RFA [6], and even CN when routing optimization [3] is possible, to register more than one CoA with each of them relating to a certain flow control policy. There are some limitations in these solutions. The methods leave the traffic control operation to the intermediate mobility agents. Each time when an end host wishes to change the policy for the traffic control, it should send new BU to all the corresponding agents. So the method is too inflexible. Policy for flow control is static and only on the basis of port number, without taking the end point s dynamic conditions into account. So the traffic control is deterministic without the necessary adaptativity. Policy Manager Connection Controller Channel APIs Traffic Analyzer Connection Monitor Protocol Entities & APIs Figure 3. Flow level traffic control The method brings more performance overhead in terms of the registration frequency, the BU extension length, and the packet processing and delivering latency. A more serious semantic problem arises. For the flow control processing in Mobile IP, network layer should concern the content of datagrams at transport level by filtering packets according to port number. End-to-end attribute of transport level is broken at mobility agents. The methods are weak to support smart applications. To modern QoS-aware applications, the connectivity situation must be known by smart applications in order to employ some adaptive algorithms. In response to these limitations, we propose a transport layer solution for the flow level traffic control, as in the next section. C. Transport Layer Solution Transport layer solutions are becoming more and more attractive over network layer solutions. This is mainly due to the changing focus from mobility management to multiconnectivity management. Since several interfaces with different IP addresses can be used simultaneously, it is too difficult to make the transport layer transparent, not to mention the more complex situation of mobility. Moreover, in a modern QoS-aware application, the connectivity situation must be known by smart applications in order to employ some adaptive algorithms. In mobility management routing is the core issue and then the network layer is the best level to solve the problem. On the other hand, data stream control, i.e. to control data packet flows of various applications to/from different network interfaces, is by nature a transport layer issue and so it should be better to implement it at least at this level. A transport layer solution is proposed as illustrated in Figure 3. It is easy to find the different effect from Mobile IP in terms of traffic control. Data stream control, i.e. to control data packet flows of various applications to/from different network interfaces, is by nature an end-to-end transport layer issue and so it should be better to implement it at least at this level. Obviously this transport layer solution is more intuitive, regular, and logical. The key issue involved in this solution is how to select the best appropriate interface as the vividly working one for each

5 connection flow among a set of available candidates. This selection is needed for both initially establishing connectivity and later triggering flow handoff in overlapping area. Criteria for the selection can be expressed as a policy, i.e. a set of rules governing decisions of the selection of network interfaces for each data flow. Both deterministic policies and conditional policies can be employed for the selection. Policy performance is depended on both static and dynamic parameters. These parameters should take the following factors into account: User context. Static parameters include e.g. user profile and preference. Dynamic parameters include e.g. user moving speed, location, time, online duration, and user behavior pattern history information. Device context. Static parameters include e.g. device profile and system configuration information. Dynamic parameters include e.g. power status and power consumption rate. Network context. Static parameters include e.g. operator and provider, network configuration, type, coverage, geographical map, typical bandwidth and latency, and charge model. Dynamic parameters include e.g. current availability status, signal strength, Signal to Noise Ratio (SNR), error and loss rates, traffic load, instant bandwidth and latency, online duration, accumulated data volume, expense, and endto-end connectivity QoS. Application context. Static parameters include end-toend architecture, service mode, security and criticality, protocol, and QoS requirements like timeliness and bandwidth. Dynamic parameters include e.g. resource allocation, process priority, data stream type, connection type and duration. These context parameters can be obtained from different sources, e.g. explicitly from hardware and software vendor, operator, and provider, or by periodically monitoring system and network at various levels. Moreover, history information is valuable for statistical analysis, estimation, and prediction. It is worth noticing that the context information is shared with other functionalities besides mobility management, in order to achieve context awareness. IV. CONCLUSIONS Mobile IP is currently the most promising solution for mobility management in mobile Internet. In this paper, two problems are distinguished as the main limitations of the Mobile IP based scheme. First, it is recognized that Mobile IP lacks flexibility with respect to the transparent and mandatory mobility support to various applications. This leads to a more adaptive architecture proposed on the basis of optional mobility mechanisms at different protocol layers. Another problem occurs when multiple interfaces are available to one user device to access heterogeneous networks. In this case flow level traffic control is needed in addition to device level mobility management as in Mobile IP. A novel architecture for multi-connectivity management is proposed to address this problem. ACKNOWLEDGMENT Financial support by the National Technology Agency of Finland is gratefully acknowledged. REFERENCES [1] C. Perkins, IP mobility support, RFC2002, IETF, Oct [2] D.B. Johnson, C.E. Perkins, and J. Arkko, Mobility support in IPv6, Internet Draft, draft-ietf-mobileip-ipv6-21, IETF, Feb [3] C. Perkins and D. Johnson, Route optimization in Mobile IP, Internet draft, draft-ietf-mobileip-optim-10.txt, IETF, November [4] R. Koodli, Ed., Fast handovers for Mobile IPv6, Internet Draft, draftietf-mobileip-fast-mipv6-06.txt, work in progress, IETF, Mar [5] H. Soliman, C. Castelluccia, K. El-Malki, and L. Bellier, Hierarchical Mobile IPv6 (HMIPv6), Internet Draft, draft-ietf-mobileip-hmipv6-07, work in progress, IETF, Oct [6] E. Gustafsson, A. Jonsson, and C.E. Perkins, Mobile IPv4 regional registration, Internet Draft, draft-ietf-mobileip-reg-tunnel-07, work in progress, IETF, October 2002 [7] A. Campbell and J.Castellanos, IP micro-mobility protocols, ACM Mobile Comput and Communications Review, Vol. 4(4), 2000: [8] Alex C. Snoeren and H. Balakrishnan, An end-to-end approach to host mobility, in Proceedings of the sixth annual international conference on Mobile computing and networking, Boston, Massachusetts, United States August 2000: [9] H. Schulzrinne and E. Wedlund, Application-layer mobility support using SIP, ACM SIGMOBILE Mobile Computing and Communications Review, Vol. 4(3), 2000: [10] P. Bhagwat, C.E. Perkins, and S. Tripathi, Network layer mobility: an architecture and survey, IEEE Personal Communications, Vol. 3, No. 3, pp , June 1996 [11] T.X. Brown and S. Mohan, Mobility management for personal communications systems, IEEE Trans. Vehicular Technology, Vol. 46, No. 2, pp , May 1997 [12] A. Acharya, J. Li, b. Rajagopalan, and D. Raychaudhuri, Mobility management in wireless ATM networks, IEEE Communications Magazine, Vol 35, No. 11, pp , Nov [13] Nikos A. Nikolaou, Konstantinos G. Vaxevanakis, Sotirios I. Maniatis, Iakovos S. Venieris, and Nicholas A. Zervos, Wireless convergence architecture: a case study using GSM and wireless LAN, ACM/Baltzer Mobile Networks and Applications (MONET), Vol. 7(4), 2002: [14] W. Mohr and W. Konhauser, Access network evolution beyond third generation mobile communications, IEEE Communications Magazine, Vol. 38(12), 2000: [15] M. Stemm and R.H. Katz, Vertical handoffs in wireless overlay networks, ACM/Baltzer Mobile Networks and Applications (MONET), Vol. 3(4), 1998: [16] N. Montavont, T. Noel, and M. Kassi-Lahlou, MIPv6 for multiple interfaces, Internet Draft, draft-montavont-mobileip-mmi-00.txt, work in progress, IETF, July 2002 [17] N.A. Fikouras, A.J. Koensgen, C. Goerg, W. Zirwas, and M. Lott, Filters for Mobile IP bindings (NOMAD), Internet Draft, draftnomad-mobileip-filters-02.txt, work in progress, IETF, July 2002 [18] H. Soliman, K.E. Malki, and C. Castelluccia, Per-flow movement in MIPv6, Internet Draft, draft-soliman-mobileip-flow-move-01.txt, work in progress, IETF, November 2001 [19] X. Zhao, C. Castelluccia, and M. Baker, Flexible network support for mobile hosts, ACM/Balzer Mobile Networks and Applications (MONET), Vol. 6(2), 2001: