A Stochastic Petri-net-based Approach For Quality-of-service Analysis Of WS-CDL-based Web Service Composition

Web services have been becoming increasingly important in the recent years. In order to ensure inter-operability between heterogeneous services across organizational boundaries, there are standard protocols which provide uniform ways to define the interface a web service exhibits (i.e. WSDL), to exchange messages (i.e., SOAP), and to look for particular services (i.e., UDDI). However, there still remain open challenges when it comes to the management of complex systems composed by a large number of services, where interactions go far beyond simple sequences of requests and responses. For this purpose, web services composition has been proposed to provide abstractions and infrastructures facilitating the definition of complex services out of simpler ones. However, research in nonfunctional properties of Web Service Choreography Description Language (WS-CDL) is very limited, especially its reliability and dependability compared with existing research on formal verification of functional properties.WS-CDL is a main-stream standard for the description of peer-to-peer collaborations for the participants for service composition. To predict the dependability of composite service processes specified in WS-CDL allows service designers and uses to decided whether the process meet the non-function requirements of trustworthiness, and to choose the process with better dependability from those with identical function. In this paper, we propose a model-driven approach for quantitative QoS(quality of service) analysis of composite service built on WS-CDL.The main innovation of this research includes a complete translation from WS-CDL to generalized-stochastic-Petri-nets (GSPN) and a dependability (using process-normal-completion-probability, expected-process-normal-completion-time and expected-overhead-of-normal-completion as the metrics of dependability of service composition) calculation algorithm based on GSPN. We also validate the accuracy of the approach in the experimental study by showing 95% confidence intervals obtained from experimental dependability results cover corresponding theoretical prediction values.