Research On Trustworthy Of Composition Web Services Based On Stochastic Petri Net

With the fast development of network technology, a new service-oriented computational model appears, which combines web technology and software engineer technology together. Accordingly, Service-oriented Architectures (SOA) come into being. SOA is characteristic of service composition and lays stress on the importance of service so that application programs can work together to provide abundant and purposeful business flows and truly reflect business models. At present, web service composition has become a hot researching field, because it can make good use of resources and thus has a promising future. In spite of the advantages of web service compostition, many issues remain to be solved in this field; the trustworthy of web service composition is one of them.The paper first introduces some basic concepts in the field of web service composition and relevant studies at present. And the paper also makes introduction to researches of the trustworthy of system. Meanwhile, three attributes of the trustworthy of web service composition are given, and they are correctness, reliability and safety respectively whose meanings are examined as well. After this, according to the purpose of the paper, both place and transition of Stochastic Petri Net are expanded, and Extended Stochastic Petri Net which has service information attached is introduced. Then the paper proceeds to introduce a kind of data place in order to analyze mistakes happening during the process of data transition among services. In terms of transition, the paper introduces two time variables to describe the rate of failure and repair of service, in order that reliability and safety of service composition can be analyzed. In the following part, the method of analyzing trustworthy of service compositon is given. With regard to correctness of models, the parameters of services are introduced into the model. Mistakes from data and their formalized description by Petri Net are presented. An algorithm which can analyze the model correctness is given on the basis of the reachability graph. As for model reliability, the average operating time of service composition is first calculated, and under the condition that the operating time is in line with exponential distribution, the probability i.e. reliability of task completion by service composition in given time. As to the aspect of model safety, failure-and-repair model is first established, and then steady probability can be worked out. In the next step, probability of supplying service can be given when service composition is attacked or network breaks down. In the last part, with the above theory, the paper makes analysis of trustworthy through an example.