          ! "# $!$ 135         Suparna Biswas 1 and Sarmistha Neogy 2 1 Department of Computer Science & Engineering, West Bengal University of Technology, Kolkata, India mailtosuparna@gmail.com 2 Department of Computer Science & Engineering, Jadavpur University, Kolkata, India sarmisthaneogy@gmail.com ABSTRACT A new checkpointing and failure recovery algorithm for mobile computing system is proposed here. Mobile hosts save checkpoints based on mobility and movement patterns.Movement patterns considered here are of three types – i) Intercell movement pattern ii) combination movement pattern ii) Intracell movement pattern. Mobile hosts save checkpoints when number of hand-off exceeds a predefined hand-off threshold value. Disconnection is a frequent phenomenon and is of two types: i) planned disconnection ii) unplanned disconnection.Hence mobile hosts save two types of checkpoints - i) permanent checkpoint based on hand-off threshold value covering unplanned disconnection ii) migration checkpoint covering planned disconnection. Hand-off threshold is a function mobility rate , movement pattern, message passing frequency and failure rate. KEYWORDS Checkpointing, movement pattern, intercell, intracell, combination, mobility, hand-off, migration checkpoint 1. INTRODUCTION A mobile computing system is a distributed system where some of the processes run on mobile hosts (MHs) moving over the network and a few fixed hosts (MSS) act as access points to communicate with MHs. Presence of the following characteristics distinguish between distributed system and mobile computing systems: • Limited Bandwidth • Limited and vulnerable MH local storage • Frequent disconnection/connection • Limited power • Cost to locate MHs Computing potential of these systems is often hampered by their susceptibility to failures. Checkpointing and Rollback Recovery is an efficient technique for providing fault tolerance to distributed as well as mobile computing systems. Mobility and frequent disconnections of MHs due to hand-off or failure is inherent in MCS. This feature of MCS and its effect on checkpointing is the prime focus of our proposed technique. Traditional checkpointing algorithms are periodic. But periodic checkpointing is not suitable for a system with mobile hosts. This is because depending on movement pattern the number of hand-offs may be more in one checkpoint interval than the other. This may result in uneven recovery time upon failures of MHs. Hence system's reliability becomes unpredictable. Checkpointing based on movement pattern and mobility rate of MHs may cause delay in checkpointing. This motivates us to introduce a concept of migration checkpoint. An MH upon saving migration checkpoint, sends it attached with migration message to its current MSS