Esearch Of Service Composition Approach Based On Reliability Evaluation

With the rapid development of computer network and the increasing popularity of Internet applications, services composition technology has been used more and more widely. In the environment of service oriented computing and cloud computing, more and more new, complex applications of service composition technology put forward higher and finer requirements. How to evaluate service credibility in dynamic distributed environment and then choose the expected optimal service execution path has not only become a new hotspot, but also been one of the problems demanding prompt solution in the entire service composition field.The main problems in current service composition field include:(1) in dynamic distributed environment, because of the uncertainty of nodes and service instances and limited communication overhead, it’s difficult to accurately evaluate the reliability of service instance;(2) in the process of choosing the optimal service composition path, current methods make little use of the service instance credibility in practice, which makes it difficult to obtain optimal results in practice;(3) most of the service composition scheduling algorithm haven’t taken the fast path recovery problem into consideration. It’s difficult to obtain the left optimal path in circumstance of execution failure.(4) In dynamic distributed environment, processes with complex structure will encounter execution problem;The main research work and innovation of present paper are as follows:1Based on the Dempster-Shafer theory, we propose an evaluation model of service instance reliability.In dynamic distributed environment, it’s difficult to accurately evaluate the service instance reliability. Based on the Dempster-Shafer theory, we propose an evaluation model of service instance reliability. Our model first calculates the direct reliability of the previous service instance to a service instance, and then synthesizes Dempster rule based on the Shapley entropy weights, and finally gets relatively accurate service instance reliability by using Pignistic probability transformation theory to subdivided uncertain interval for reliable evaluation. The experimental results indicate that in the dynamic environment the service instance credibility evaluation model is effective, and the reliability assessment of the value of accurate rate is90%above. 2we present a credible service scheduling method.In dynamic distributed environment, most of the optimal service composition paths calculated by the current service scheduling algorithms can’t obtain the actual effect at the time of executing. Aiming at this problem, in this paper, based on the service instance trust evaluation model, we propose a credible assessment based service scheduling algorithm. Our algorithm first computes accumulation of the expectation polymerization property of all alternative paths, and then selects the path with a minimum expected aggregate attribute accumulation as the optimal path. In dynamic distributed environment, it is difficult to accurately model the system, thus affect the optimal path calculation. This article proposes a service combination algorithm based on the Q learning. This algorithm first gets a rough Q initial value of the combination of environment by simple modeling, and then through the feedback learning, constantly iterative Q, in order to obtain more accurate combination strategy. Experiment results show that in dynamic distributed environment, the combination success rate and the comprehensive performance of QoS of the two algorithms have been effectively upgrade. Compared with traditional service combination algorithm, the comprehensive performance of QoS of our credible assessment based service composition algorithm has increased by15.2%, and the success rate has increased by22%.3We propose a credible service combination method which supports dynamic local optimal.In according to the execution failure of optimal path and difficulty to obtain the optimum routing problem, this paper proposed service combination algorithm based on the local state reverse diffusion. This algorithm uses decision diagram optimization idea, by periodically reporting QoS state from back to front, we can calculate the optimal and suboptimal paths deployment step by step. In practice, if the optimal path on a service instance fails, we can directly choose suboptimal paths as the optimal path, which ensures the execution of rapid recovery. In distributed environment, not all branches of branch convergence flow can convergent to one execution node. This paper presents a decision algorithm based on common node branch convergence, which makes convergence process of branches smoothly executed. Simulation and practical experiments verify the validity of the algorithm, which can guarantee the optimum and optimal local requirements.