Load management in model-aware execution of composite web services

Abstract

In the Service Oriented Architecture services are computational units that can be published, discovered, consumed and aggregated in the platform and organization independent manner. The most widely accepted way to achieve Service Orientation (SO) is with Web Services (WSs), due to the standardization efforts and the wide range of available infrastructure. One of the most interesting aspects of WSs is the ease with which they can be combined into Composite Web Services (CWSs). The most popular language to specify and implement CWSs is BPEL. While being easy to use, it also introduces difficulties to monitor and optimize CWSs, specifically in the selection of optimal WSs. This paper investigates the possibility to support this selection with dynamic load management, based on the alternative, model-aware, approach to orchestrate WSs with the Coloured Petri Nets (CPN) formalism. The use of the mathematically grounded formalism allows to model and verify properties of CWSs and enables at runtime guidance of the execution of the CWS. This paper presents how, during a model-aware execution of a CWS, to predict and avoid some of the undesirable behaviors of WSs. Compared to BPEL, the model-aware approach significantly improves the performance and manageability of CWSs and thus opens up new deployment scenarios.