Efficient traces' collection mechanisms for passive testing of Web Services

Web Services are a novel approach for business-to-business interactions. Their management, especially fault and performance management, is becoming necessary for their success and emergence. Nowadays, this management is platform-dependent and does not allow third parties to be involved. In this paper, we consider management of Web Services by passive testing where the tester itself is a Web Service. We propose different architectures for observation of simple and composite Web Services. We also study a set of online traces collection mechanisms and discuss their performances in terms of required CPU/RAM resources and introduced network overhead. These performances are then maximized by selecting best locations of observers. Observation considers both functional and non-functional (QoS) properties of Web Services. The paper presents also our experiments using different observation architectures and traces collection mechanisms while observing a simple and a composite Web Service.