Research And Implementation Of Monitor And Diagnosis Framework And Algorithm Based-on SOA

Service-Oriented Architecture (SOA) is an effective way to control cost when building multi enterprise application. Using SOA, enterprise systems can define and execute transactions that across multiple server domains at distributed locations. However, The manageability of large-scale business processes are challenging. It is important to have a mechanism to identify services'responsibility for end-to-end QoS failure at run-time, including both service level agreements (SLAs) violation and functional errors. Therefore, the flexibility and dynamism of SOA makes it critical to monitor and manage services behaviors at runtime for performance assurance.Firstly, this article elaborated the related theory knowledge about SOA and Web Service, introduced technical concepts of fault diagnose using Bayesian network reasoning:including accountability framework,sensitivity analysis technology,k-median problem,diagnose parameter. The article presented accountability network fault diagnose framework in details, and the related Bayesian network diagnose algorithm. The article composited Bayesian sensitivity analysis technology and k-median model, improved fault evidence set selection algorithm and fault diagnose algorithm by adding fault identify set, in the basis of original ECS fault evidence set selection algorithm.The main task of the thesis is the design and simulation of monitor and diagnose environment that based on SOA, added more details about accountability monitoring framework, also improved Bayesian network monitoring and diagnose algorithm, that is, adding fault evidence set,fault event set and fault identify set:firstly, lessen the range of service that needed to be diagnose by using improved fault evidence set algorithm before Bayesian reasoning process, also recorded history data of fault event; specified various functional component of accountability monitoring framework, applied improved network diagnose engine in this framework. Through the experiment, compared performance of using improved fault evidence set selection algorithm and ECS algorithm. Top-cluster channel selection,bottom--cluster channel selection algorithm to execute diagnosing, the superiority of the improved fault evidence set selection algorithm proposed by the article is verified. Based-on event-driven evidence strategy used by the article, enhanced the performance of fault diagnosing significantly, and also has advantage to the future's predictive analysis and evaluation of history reputation.Finally, the thesis summarizes the diagnosing strategy,design of simulation environment and algorithm based on SOA network monitoring framework, as well as the direction for further research and improvement.