A UML profile for quality-based web service selection using logic scoring of preference method

Abstract

In the modern information infrastructure, Web services have a very important role. Web services are one of the main tools for information creation, delivery and processing in society and economic life; for the formation and further evolution of the information space. Among the existing and constantly growing number of Web services, in a natural objective way, there are many Web services with the same or similar functional properties. This makes it possible to select among them such Web service that best satisfies the consumer’s requirements and is most suitable to consumer’s preferences on quality characteristics, for instance, cost, reliability, response time, etc. On the other hand, the consumer is not always possible to decide which Web service is the best from his point of view, because Web service has multiple quality characteristics. In this regard, many researchers formulate the problem of quality-based Web service selection as a multi-criteria decision-making problem and integrate the facilities for the decision-making into decision support systems. One of the methods of multi-criteria decision-making is the Logic Scoring of Preference (LSP) method. Given the current trends in the transition from document-based to model-based systems engineering, it is expedient to provide the design templates (meta-models or profiles) for the decision-making support systems developers. In this paper, a UML profile to design the quality-based Web service selection systems using the LSP method was proposed. Standard UML extension mechanisms (stereotypes, tag value definitions, and constraints) to capture the basic quality properties of the Web service and the rules of applying the LSP method operators for this profile were used. The efficiency of the proposed profile by the examples of modelling the LSP aggregators to aggregate the Web service quality characteristics was demonstrated. The correctness of the models created using the proposed profile in the MagicDraw modelling environment was verified.