Metrics, Methodology, and Tool for Performance-Considered Web Service Composition

摘要

Web service is a promising technology to efficiently integrate disparate software components over various types of systems and to exchange various business artifacts across business organizations. As many web services are available on Internet, performance becomes increasingly important to distinguish different service providers. Performance is a difficult property to assess, since it involves non-deterministic networks, frequent changes on workload and unexpected usage patterns. In this work, we propose a performance analysis methodology for web services, and introduce sPAC (Web Services Performance Analysis Center) that implements our methodology for performance-considered web service composition. sPAC 1) graphically describes web services and the flow between them, 2) automatically generates test codes for the web process and executes them for performance analysis under low load intensity, 3) automatically generates a simulation model for the web process, and conducts extensive simulations for heavy load intensity, and 4) reports analysis and estimation results to help service customers verify timeliness of their web services.