The Implementation of A Dependency Matrix-based QoS Diagnosis Support in SOA Middleware

When an SOA business process fails to deliver the desired quality of service (QoS), it is necessary to identify the faulty services that cause the problem since the source of the problem may not be at where the problem is observed. In this paper, we propose a polynomial time diagnosis algorithm by using a dependency matrix for business process structure in SOA. The dependency matrix is built based only on process workflow structure, with no need for historical knowledge on prior executionx. By comparing the performance data reported from business process probes, the proposed diagnosis algorithm also checks some predicates-on-probes (PoP) to increase the monitoring and diagnosis accuracy. We have implemented the diagnosis support for the dependency matrix based QoS management in the Llama middleware. A performance study using some realistic services running on networked Web servers shows that the system can achieve a diagnosis completeness of up to 80%.