Policy-based runtime performance evaluation and validation of web services

Service-Oriented Architecture (SOA) enables the dynamic composition of services across heterogeneous platforms through standard Internet protocols. However, it is hard to evaluate service runtime qualities due to the instability and uncertainty of the open Internet environment. Testing is usually exercised in the dedicated testing environment and cannot analyze service behavior in the execution environment. Hence, runtime monitoring is necessary for runtime system status tracking and anomaly detection. The paper is an extension of our previous research on the collaborative monitoring of service-based systems. It defines a service performance evaluation model using the temporal properties of both atomic and composite service executions. Policies are defined and encoded following the WS-Policy framework. They are enforced at runtime for evaluating the service performance based on the evaluation model. A prototype system is implemented. Experiments are exercised to illustrate the proposed approach and analyze the monitoring overhead.