Abstract:

In the process of increasing the number of work-flow services, as the potential cost of the initial planning is high, and the reliability of many service composition algorithms is not enough. A service composition algorithm based on the location of backup service and probabilistic quality-of-service（QoS）model was proposed. A service cluster based QoS optimization selection for each service was computed which contained a sufficient number of backup services. The location of backup service was considered to be evaluated at failure of a single point. Because the backup service was distributed evenly, the occurrence of the failure of the task was presented. For the service selection problem, an improved multiple objective optimization（MOO）algorithm was adopted to calculate the feasible solution set using clustering and QoS model. JMETAL 3.1 framework was used to assess the returns of the backup service location and the reliability of the algorithms. The results show that compared to other MOO algorithms, the proposed algorithm is more reliable and the gain from the backup service position is higher.

Key words: workflow, service composition, probabilistic quality-of-service, multiple objective optimization, clustering