Research On Non-Blocking Coordinated Checkpointing Algorithm In The Mobile Computing Environment

The mobile computing system has the extensive development and application prospect, due to its characteristics including dynamic mobility, flexibility, autonomy etc. However, it is inevitable to reduce the dependability because of the unreliability of mobile devices and the vulnerability of wireless link connection. Checkpointing and rollback-recovery technology is a low-cost software fault-tolerance technology in the traditional distributed system. However, it could not be applied to mobile computing systems directly by reason of the introduction of many new problems in mobile computing environments. So, in allusion to the new characteristics of mobile computing environments, it is a significative work to design checkpointing and rollback recovery algorithm for mobile computing. It is very important to promote the use of the mobile computing system in high-performance computing and other fields.There are many advantages of coordinated checkpointing algorithm in traditional distributed systems, but its larger synchronization overhead makes it not suitable for mobile computing environments. Therefore, in order to reduce overhead, save the wireless network load, it designs the non-blocking coordinated checkpointing algorithm that based on mobile computing environment. The completion of checkpoint initiated and coordinated are both on the mobile support station, then through mobile support stations are time-based synchronization. The algorithm is transparent and low-cost for mobile host,and can rollback to the consistent and resumable state by avoiding the orphan messages, saving the in-transit message and non-blocking the basic implementation of the calculation process.To further enhance the efficiency of checkpointing algorithm, it improves the storage strategy of checkpoint in mobile computing environment. In allusion to different types of failure on mobile host, checkpoints are stored in local disks of the mobile host and the stable storage of mobile support station separately. By using low-cost checkpoints to tolerance high-frequency fault, and high-cost checkpoints to tolerance little fault, the load of wireless network and the delay accessing checkpoint both in the fault-free execution and recovery are reduced, thereby the energy consumption is also reduced, and it is available to the minimal average cost according to dynamic configuration algorithm based on the actual environment.It makes performance evaluation about the proposed algorithm. In the first place, it proved the correctness, transparency and low cost about non-blocking coordinated checkpoint algorithm; in the next place, it established a performance evaluation model using Markov process; in the end, it designed simulation experiments to verify the superiority of the algorithm by numerical way.