Research On Checkpoint Storage Scheme In The Mobile Computing System

Compared to the wired distributed computing system, certain new characteristics such asthe small storage space and the limited electric energy of mobile host, easily being disturbedand the limited bandwidth of wireless net make the processes on mobile hosts in mobilecomputing system fail more easily. If there are no steps taken to deal with the failed mobilehost processes, the failed mobile host processes will do great harm to the system.Fault-tolerant technology is an effective measure to guarantee the mobile computing systemrun normally. Checkpoint technology is one kind of fault-tolerant technology. The storage ofthe status of mobile host processes which is called checkpoint information is a problem ofcheckpoint technology. Whether storage strategy of checkpoint information is reasonableaffects the performance of system.In this paper the average time when a process completes a standard computing task istaken as an important indicator of the performance of system. The average time when aprocess completes a standard computing task increases with the increase of the average extratime. System usually adjusts the position of checkpoint information when a mobile hosttraveling from one area to another area. There are three kinds of traditional checkpoint storageschemes including Eager Strategy, Lazy Strategy and Trickle Strategy. But they migratecheckpoint information blindly and pull in high average extra time spending during a processcompletes a standard computing task.This paper analyzes the advantages and disadvantages of several traditional checkpointstorage schemes and introduces a new checkpoint storage scheme called Strategy Based onmobile host Process Failure Probability by inheriting traditional strategies’ advantages andgetting rid of traditional strategies’ disadvantages. This strategy takes the stability of mobilehost processes as the basis of whether to migrate the checkpoint information to avoid theblindness during migration. It can guarantee less average extra time spending under all failureprobability of the mobile host process so as to reduce the average time when a processcompletes a standard computing task. This paper proposes a failure probability model of themobile host process which is used to calculate the failure probability of the mobile hostprocess. So system can judge the stability of the mobile host process when the mobile hostmoving from one area to another. Finally, this paper carries out theoretical analysis of these strategies and verifies theresults of the analysis by experiments. The results show that Strategy Based on mobile hostProcess Failure Probability can complete a standard computing task using less average timeand the performance of mobile computing system has been improved.