Research On Asynchronous Checkpoint Algorithm Based On Mobile Wireless Networks

With the popularity of notebook computers, smart phones, PDA and other mobile computing devices, the construction of mobile wireless networks is rapidly developed, and the types of wireless networks are also increasingly diverse. Checkpoint algorithms based on mobile wireless networks become an important branch of fault-tolerant computing. Because different types of wireless networks have their own characteristics, it is not suitable that traditional checkpoint algorithms transplant directly to the mobile wireless networks. Therefore, the research for checkpoint algorithms based on mobile wireless networks has great significance.Checkpoint algorithms are divided into the algorithms which are depended on checkpoint and the algorithms based on the message logging. The algorithms which are depended on checkpoint include the following categories: synchronization algorithms, asynchronous algorithms and signal-induced algorithms. Checkpoint algorithm based on message logging has advantages that output fast, checkpoint flexible. Because of the characteristics of mobile wireless networks, the method of combining algorithms based on message logging and checkpoint technology has a good application prospects.This thesis researches the characteristics of mobile wireless networks, introduces some basic concepts, development and applications about checkpoint technology, analysis of the current strategies of checkpoint algorithms, and the impact of checkpoint technology for mobile wireless networks, and presents a kind of asynchronous algorithm based on 802.11 network protocol. Algorithm uses two-stage message logging. In this thesis, in the condition of failure meet the Poisson distribution, proposed a kind of dynamic checkpoint strategy based on exponential distribution and normal distribution, and proved that the checkpoint interval coefficients and the value about system failure meet the exponential distribution. Comparing the simulation shows that the algorithm proposed in this thesis have low cost, is more suitable for mobile wireless networks with limited resources.