Reliability of SOA Systems Using SPN and GA

The aim of the paper is to develop a methodology for reliability data estimation of services, and to develop a model to select the most suitable set of services for service composition, to support the reliability requirement of building a SOA system. The proposed methodology involves qualitative modeling of service composition using Stochastic Petri net (SPN) and quantitative analysis using fuzzy set theory for representing the failure rates and repair rates. Failure rates and repair rates are incorporated into SPN model and system reliability parameters are computed. A multi-objective genetic algorithm (GA) is adopted to select the most appropriate set of services using reliability parameters such as availability, reliability, and MTBF as the criterion of interest. The adopted methodology gives a graphical representation of the structure of the system, and selects the most reliable services. An example of a travel process has been performed to demonstrate the proposed method. The results may be helpful for the designers to assess the reliability data of the system and to select best services for building the more reliable SOA system.