A Performance Study on REAchability Protocol in Large Scale IPv6 Networks Habib Naderi Department of Computer Science University of Auckland Auckland, New Zealand e-mail: hnad002@aucklanduni.ac.nz Brian Carpenter Department of Computer Science University of Auckland Auckland, New Zealand e-mail: brian@cs.auckland.ac.nz Abstract-- Shim6 is a host-centric multihoming solution for IPv6 networks which has been chosen by the IETF as a multihoming solution for the future Internet. Shim6 employs REAchability Protocol (REAP) for failure detection and recovery. REAP enables multi- addressed hosts which are involved in a communication to detect and recover path failures without breaking the transport layer sessions. One concern about deploying shim6 is the performance of REAP in large networks like campus or enterprise networks. We modeled and simulated this protocol by using stochastic activity networks (SAN) and performed an analysis of its performance. This paper presents the results of the simulation. We also investigated the effect of send timer and initial probes on the performance. Keywords: IPv6, Multihoming, Shim6, REAP I. INTRODUCTION Multihoming is one of the desirable features for the current and future Internet. Fault tolerance, which is provided by multihoming, enables businesses to offer a continuous service to their customers and have the opportunity to compete with their rivals. In the current Internet, most large sites use BGP and its features to achieve this functionality. But this solution suffers from a big drawback: scaling. This solution inserts new entries to global routing tables and can lead to routing table explosion. So, in recent years researchers have been busy seeking a new solution for the future Internet using IPv6. Different solutions have been proposed for this problem [1]. Shim6 [3] is one of those solutions. Shim6 is a host-centric multihoming solution for IPv6. In this category of solutions, hosts are responsible for providing multihoming service and the routing system is unaware of this functionality. This category of solutions is valuable because deployment would be highly incremental. Shim6 inserts a new layer inside IP. This layer performs a mapping between identifiers (ID) and locators. This mapping is transparent to the transport layer, so shim6 can switch between locators without breaking a transport session, another desirable feature of multihoming [2]. One important part of shim6 is the algorithm which is used for failure detection and recovery. Shim6 employs a separate protocol called REAchability Protocol (REAP) [4] for this purpose. REAP monitors transport layer communications and when a failure happens, it tries to recover. Recovery is performed by sending special messages called probe messages to the other end of communication to find a new working pair of available locators. One probe for each address pair is sent until a proper response is received from the peer. The new address pair is then adopted by shim6 without impact on the transport. The Internet Engineering Task Force (IETF) chose shim6 as a solution for multihoming in IPv6 networks and the future Internet. Performance is a key feature for new Internet protocols. Shim6 should be able to present an acceptable performance, especially in large scale networks with thousands of computers, like campus and enterprise networks. Antonio de la Oliva, et al. [5] performed a performance analysis on REAP to evaluate its behavior, but with just two hosts, each with two addresses, two TCP and two UDP based applications. Our results match their results although the aim of their work is different from ours. In this paper we present the results of a performance study on REAP in large scale networks. We modeled REAP with Stochastic Activity Networks (SAN) [6] and used the Möbius modeling tool [7] to simulate the model. Our results show the behavior of the REAP in a network with thousands of connections. The remainder of the paper is organized as follows: an overview of REAP is given in section II. Section III presents the simulation environment and our assumptions. The results are presented and discussed in section IV. The effect of the send timer and initial probes are discussed in section V. We conclude our work in section VI. II. OVERVIEW OF REACHABILITY PROTOCOL Shim6 employs a protocol for failure detection and recovery called REAchability Protocol (REAP). REAP is responsible for two main jobs: detecting failure and finding another operational address pair,