Research Of Fault Tlerant Of SOA In The Field Of Performance Based On Queueing Theory

This paper is supported by the National High-Tech Research Development Program of China (863 program) under Grant No.2007AA01Z139, and research on fault tolerance of service performance,the study includes:First of all, this paper discusses the definition of the fault of service performance. And then this paper proves that the service fault detection is not only an important part of fault tolerance technology in SOA applications, but also a vital part of dynamic discovery and composition of services.Secondly, this paper proves the correctness of using queueing theory to solve performance problems in SOA applications. We find out that the service requests processing has some features which are accord with the assumptions of the queueing models.Thirdly, the performance of services of SOA applications is measuring in two categories (individual services and composite services). In this paper, the author uses the queue model M/M/S/K to analysis the individual services and in the case of composite services, it is solved using the mean value analysis algorithm for multi-class closed system. As we analysis the performance problem in two case, we get a fine algorithm of performance evaluating.Fourthly, this paper discusses the detailed implementation of the prototype system of the model and the rational of the mechanism. First we discuss how to collect data to compute the mean arrival rate and mean departure rate of service. We use the third party software to intercept at network layer and write the datum into database. In this process, we differentiate the services by URL as unique identification. After computing of the mean arrival rate and the mean complete rate, we pass these values to the service performance validation components as arguments and finally analysis the performance of the service.And then, this paper also discusses the mechanism of service fault detection, fault tolerance and future works.This thesis puts forward a mathematic model based on queueing theory which can dynamically evaluate the performance of services and extend the idea of research in QoS research area. It also laid the theory foundation of dynamic service discovery and composition for SOA frameworks.