SOA-based method for cooperative tasks distribution in large-scale optimization environments

In this paper, operators are encapsulated by services in algorithms for large-scale optimization problems and the services are deployed in distributed systems. Response time is an important factor for the performance of cooperative services. The message parse time is analyzed. Initial service accessing time is the main overhead for service calling. To decrease both the initial time and the response time, SBP (Service Buffering Pool based Scheduling Algorithm) is proposed by integrating a resource pool with a cache. Appropriate number of computing resources can be determined for large-scale optimization problems after analyzing the influence of the number of computing resources on algorithms. The proposed algorithm is compared with a centralized algorithm and a distributed algorithm without cache. Experimental results show that the proposed algorithm has a lower overhead than the other two algorithms.