A model-driven approach for QoS prediction of BPEL processes

Abstract

Business Process Execution Language (BPEL) is one of the most popular languages for Web service composition. To predict the QoS of composite service processes specified in BPEL gives the way to tell whether the process meet the non-function requirements, and to choose the process with better QoS from those with similar function. A model-driven approach for QoS prediction of BPEL processes is proposed in this paper, which has a two-layer architecture: One is the prediction model specifying necessary information for prediction and independent of specific languages, and the other is the semantic model of specific languages such as BPEL. A set of transformation rules is defined between the two layers so that processes specified in specific languages can be transformed to the prediction model. A prediction algorithm is defined based on the prediction model, and through the algorithm the average value of process QoS attribute can be computed. The approach can be used not only to BPEL processes but also to processes in other specifications such as BPML and BPSS, if the mapping rules between the semantic models of these languages and the prediction model are defined. The feasibility and good accuracy of the approach has been proved by the experiment. © Springer-Verlag Berlin Heidelberg 2007.