Composite Service Reselection Method in Correlation Context

Abstract

Reselection of composite service is one of the core research issues in service computing field. In practical scenarios, because of business relations between service providers or the restrictions of service physical deployment environment, there may be correlation among services. Its existence affects the quality of services used together, makes corresponding abstract services potentially correlating with each other. This leads the QoS used to determine the bindings between abstract and concrete services to be inaccurate, and the selected services would not be the optimal one in actual execution environment. In this paper, the QoS-Correlation services are extracted from the execution log through Apriori data mining method. Then, the authors capture the correlated abstract services to present such correlations in a higher level of abstraction regardless of actual services. In the final, the correlated abstract services will be regarded as a task unit, and the corresponding correlation services of each task unit as its candidate services set. They also propose a correlation-aware service selection method in the paper. The method includes runtime reselection whenever the actual QoS largely deviates from the estimates.