Predicting And Analyzing Performance For Web Service-Based Cross-organizational Workflow

The development of agile supply chain and economic globalization promotes that enterprise integrated information system has been extended from internal to cross-enterprise alliance. Providing organizations with an environment to enhance B2 B communications, the SOA paradigm creates modularized services supporting the business processes within their organization and also those external entities participating in these same business processes. Web services are an increasingly important and practical instance of SOAs, which makes Web service-based workflow come true. The business process needs reaching a certain performance indicator to achieve business goal. It is of significance to effectively predict the workflow performance and analyze its relation with Qo S of each sub-service in the composition services.In the environment of service computing, different service users may experience quite different performance on the same Web service. Therefore, this paper proposes an approach for Web service-based workflow performance prediction and analysis from users’ perspective. The main work in this paper is as following.Firstly, research the theory of Web service-based cross-organizational workflow, propose the whole process of monitoring, predicting and analyzing workflow performance from users’ perspective.Secondly, through analyzing users’ characteristic and services’ characteristic, propose an evaluation approach for Qo S time series based on singular value decomposition, and calculate workflow performance time series according to the Qo S time series.Thirdly, adopt Neural Networks model trained by workflow performance time series above to predict workflow performance. When the predicting value is beyond or below preset threshold, analyze and find the key sub-service which causes performance violation.At last, conduct simulation for the proposed approach. The experiment result showed the effectiveness and advantages of the Qo S time series evaluation approach and workflow performance prediction approach, and the rationality of workflow performance analysis approach.