Second-Order HMM-Based Diagnosis Method for Web Service Fault

Web service application is becoming a popular and important software application on the web. With the advent of more and more service application using on the web, it is a crucial challenge to use an effective diagnostic technology to localize the service faults. To improve the diagnosis capability for service application software, we build a second-order hidden Markov diagnosis model by considering the dependency relationship of three ordered behaviors in the service process. Furthermore, we extend the Viterbi algorithm to find the most likelihood correct behavior sequence with the exception trace using the second-order HMM we propose. Finally, we compare the correct sequence with the exception one to localize the service faults. Experimental results show that our method is effective to the web service application in the fault diagnosis.