PEPA Analysis of MAP Effects in Hierarchical Mobile IPv6 Hao Wang and Dave Laurenson Institute for Digital Communications, Joint Research Institute for Signal & Image Processing, School of Engineering & Electronics, University of Edinburgh Email: {H.Wang, Dave.Laurenson}@ed.ac.uk Jane Hillston Laboratory for Foundations of Computer Science, School of Informatics, University of Edinburgh Email: Jane.Hillston@ed.ac.uk Abstract—To overcome the drawbacks of the Mobile IPv6 protocol on handling local mobility management, IETF proposed the HMIPv6 protocol which introduces an intermediate mobility anchor point (MAP) to hide the movement of a mobile node within a local area. However, the MAP forms a bottleneck in the network since all the traffic destined for its served nodes has to go through it. Most research on HMIPv6 focuses on protocol optimisation, and performance analysis of HMIPv6 is usually simulation-based. In this paper, we employ a performance evaluation formalism named PEPA to investigate the performance tradeoffs of MAPs in HMIPv6. Performance measures such as response time and MAP utilisation are presented. I. I NTRODUCTION To provide continuous connectivity when mobile users change their points of attachment to the Internet, the IETF proposed mobility management protocols Mobile IPv4 [1] and Mobile IPv6 [2] to support global mobility in IP-based networks. In Mobile IPv6-aware networks, a mobile node is always addressable at its home address regardless of its location. Whenever a mobile node moves into a new access network, it acquires one or more care-of addresses representing its current network attachment. The mobile node needs to send Binding Update messages (BUs) which associate its home address with its current care-of address to the mobile node’s home agent (HA) and all the correspondent nodes (CNs) it is communicating with. The movement of the mobile node can then be made transparent to the transport and higher-layer by mapping home address to care-of address at the network layer. However, although the Mobile IPv6 protocol supports a route optimisation communication mode, the quality of service will decrease if the mobile node changes its point of attachment so frequently that handoff latency and signalling load caused by Binding Update messages become significant. To overcome this drawback of the global mobility manage- ment protocols, IETF proposed local mobility management protocols such as Cellular IP [3] and Hierarchical Mobile IPv6 (HMIPv6) [4]. The HMIPv6 minimises the amount of signalling outside a local domain by using a new mobility agent, called a Mobility Anchor Point (MAP), that can hide the movement of the mobile node within a local domain. However, the MAP has to operate as a relay node between the mobile node and the CNs since by design all the traffic must go through the MAP. Under heavy traffic conditions, this local mobility management results in the MAPs becoming the bottlenecks of the network and thus network performance is degraded. In this paper we use a performance evaluation formalism named PEPA to investigate the effects of MAPs on the response time and MAP utilisation in HMIPv6 with a client- server architecture. In particular we investigate the number and placement of MAP nodes within an access network. The rest of paper is organised as follows. In Section II we introduce the PEPA formalism. The HMIPv6 protocol is reviewed in Section III. We present our PEPA model of HMIPv6 and derive performance measures in Sections IV and V respectively. Section VI presents our conclusion. II. PEPA Performance Evaluation Process Algebra (PEPA) [5] is both a timed and stochastic extension of classical process algebra such as CCS [6] and CPS [7]. In PEPA a system is described as a component or a group of components that engage in activities. Generally, components model the physical or logical elements of a system and activities characterise the behaviour of these components. Each activity a in PEPA is defined as a pair (α, r) — action type α and activity rate r. The action type can be regarded as the name of the activity and the rate specifies the duration of the activity which is an exponentially distributed random variable. If a component P behaves as Q after completing activity a, then we can denote this transition as: P α −→ Q or P (α,r) −→ Q The PEPA formalism provides a small set of operators which are able to express the individual activities of com- ponents as well as the interactions between them. We only present the operators we used in our model in this section. For more details about PEPA operators, see [5]. Prefix: (α, r) .P The component (α, r) .P carries out an activity that is of action type α and has a delay that is exponentially distributed with rate r, which gives an average delay of 1/r. After