Performance Optimization For Composite Web Services Based On Decentralized Control Mechanism

Web service which has low coupling, reusability, platform independence and composition, provides a perfect solution for the integration of cross-platform application system within one enterprise or among different enterprises. However, in order to meet the increasing complexity and ever-changing business requirements, most of the single web services have been composed to form a value-added web service composition. A centralized execution mode is used in typical composite web services, and the solution of centralized execution will not only cause unnecessary network traffic and data congestion, and result in immense communication overhead, but also lead to more limitations like operating efficiency, applicability and scalability on the web services system.By analyzing the correlative technologies and solutions in decentralized execution mode of composite web services, this paper proposes a solution of performance optimization for composite web services based on finite state machine in order to overcome the limitations of centralized execution mode. This solution uses the way of state transition to represent the finite state machine, which adopts heuristic searching algorithm in state space to execute the path which has the global optimum overhead. This solution can effectively reduce the number of passing messages and improve the throughput of system, and achieve the functions of composite web services by communication between the various states. Finally, QoS performance evaluation model which evaluate the communication cost between the various states is built to evaluate execution performance of composite web services with this solution. In order to illustrate that decentralized control mechanism has the better execution performance compared to the centralized control mechanism, this paper applys the solution of performance optimization to Huamao Barter Trading System. This system designs the business logic system which is composited of various subsystems into a web service. The centralized control mechanism is compared to the decentralized one in this paper, and the cost of the system is evaluated in the paper. With QoS performance evaluation model, the experiments show that the solution of decentralized control mechanism plays a very good performance in optimizing this system.