A Multiple System Performance Monitoring Model for Web Services

With the exponential growth of the world wide web, web services have becoming more and more popular. However, performance monitoring is a key issue in the booming service-orient architecture regime. Under such loosely coupled and open distributed computing environments, it is necessary to provide a performance monitoring model to estimate the likely performance of a service provider. Although much has been done to develop models and techniques to assess performance of services (i.e. QOSs), most of solutions are based on deterministic performance monitoring value or boolean logic. Intuitively, probabilistic representation could be a more precise and nature way for performance monitoring. In this paper, we propose a Bayesian approach to analyze service provider's behavior to infer the rationale for performance monitoring in the web service environment. This inference facilitates the user to predict service provider's performance, based on historical temporal records in a sliding window. Distinctively, it combines evidences from another system (For example, recommendation opinions of third parties) to provide complementary support for decision making. To our best of knowledge, this is the first approach to squeeze a final integrated performance prediction with multiple systems in Web services.