Research On Runtime Resolution Of Feature Interaction For Composite Web Services

Feature Interaction problem originates from telecommunication domain and firstly studied by Bellcore. However, this phenomenon is not unique to the telecom, but in some software system beyond telecom. In web service domain, because of the openness and changeability of Web service and Internet, new services are developed and deployed in the Internet; it also brings the potential of the problem resembling the feature interaction in the telecommunication area, which negatively impact service quality and user satisfaction. Such undesirable interactions have been studied as the Feature Interaction Problem(FIP for short) in Web service. On the one hand, FIP negatively impact service execution, reduces service development efficiency and decrease reliability. On the other hand, FIP results in decreasing the rate of customer’s satisfaction. Therefore, maintaining consistency of Web Service in the business process is a concern to be addressed. What’s more, FIP always occurs in business process on account of the diversity and personality of services. Therefore, online management of FIP in web services should be concerned with.In this paper, we concentrate on the online resolution of FIP using bi-objective programming theory based on the detection work. Assuming that FIP has been detected by the detection method, and therefore, the main work of the paper is resolving the FIP without changing user’s initial requirements. The problem is studied from the perspective of user’s revenue, and an online approach to FIP is proposed based on our previous work. First, the FI problem in web services is transformed into a bi-objective optimization with constraints, and then the feature interaction bi-objective optimization model is established to get a set of Pareto solution; second, the Pareto solution is ranked to identify the optimal one, which includes the right rollback strategy and a new plan. The case study illustrated that our approach is efficient to solve the Web service feature interaction problem at runtime in viewpoint of the users’ revenue.