Research On The Technology Of Fault Detection Under The Mobile Computing Environment

With the development of information and network technology people made the request of accessing to information anytime and anywhere. Mobile computing that based on distributed computation and wireless communication came into being. Mobile computing network has the characters of narrow bandwidth, communications vulnerable to interference, dynamic topology, mobile hosts often disconnected from the network, and mobile nodes by the software and hardware resource constraints. It is more vulnerable than the traditional distributed computing systems. The calculation error and performance bottlenecks caused by a verity of failures become the huge obstacle of popularizing application of mobile computing technology. Furthermore, the mobile computing, as a new distributed computing technology, has some otherunique characters. So the fault detection methods for traditional distributed computing systems can not be directly applied in the field of mobile computing. Because of which fault tolerance and recovery strategy designed for mobile computing environment lacks actual theoretical and technical support.The technologies of fault detection that is designed for traditional distributed computing systems, were in-depth studied in the paper. Among the technologies, more attention was paid on the comparison-based one. In addition, some kinds of comparison-based fault detection strategies devised for mobile computing systems were considered. Combined with the characteristics of mobile computing networks, the key problems required to solve before using comparison-based fault detection techniques under mobile computing environment were analyzed. A mutual-test-comparison-based fault detection algorithm for detecting fault nodes in mobile environments which were built on cellular networks, was proposed in the thesis. And it is an improvement from Generalized Comparison Model, according to the characters that Cellar networks have a Base Station(BS) as the central node with high reliability in every cell. Firstly, to take advantage of these characters, the algorithm uses mutual test-comparison to build a "Symptoms Chart", and divides into two steps. In the first step called detection step, every mobile node test the BS and the BS also tests every mobile nodes. This is to let all fail-free mobile hosts involve in, so as to avoiding the situation that all mutual-test-comparison results invalid when the BS is fault, to improving the reliability of mutual-test-comparison results. Secondly, the algorithm makes use of the results generated in the first step and compares if the results set keeping statistic regular, which is to verify whether the data generated in first step was credible. There are four situations judged by the algorithm, each one match a procedure. And after executing the one of four procedures all fault nodes can be identified. At last, some strategies for resolving dynamic network topology were proposed.The feasibility and performance of the algorithm were examined from both theoretical analysis and simulation. All results show the algorithm is suitable for the cellar network mobile computing environment. And it has high diagnosability but low overhead on executing and communication.