Abstract: With the wide applications of Web services and composite services in the distributed network,the size and complexity of Web service are increasing continuously.These could cause various faults of service system during the running.In building high-reliable service applications,one of the critical challenges is how to localize faulty service quickly and exactly and help service engine restore the normal process as soon as possible.To perfect the diagnosis mo-del and minimize the impact of noise data on diagnosis accuracy,we presented a service behavior model-based diagnosis method of behavior inference.The method models hidden markov model by combining historical data into service process definition.On the basis of using the decoding algorithms in HMM,the method is able to infer a correct execution trace which has the maximum likelihood with the exception execution trace and localize the service faults by comparing the differences between them.The experimental results show that the method is effective and robust to various noises in diagnosing the faults of Web services.

Key words: Web service,Model-based diagnosis,Hidden markov model,Noise data,Historical data,Service process