Research On Distributed Service Discovery Based On Grey Wolf Optimization Algorithm

In the large-scale distributed dynamic heterogeneous service computing scenario,the addition and withdrawal of services will become very frequent,and the structured P2 P network will continue to restructure,which greatly affects the response time and recall rate of service discovery,and thus greatly affects the overall availability of the system.Therefore,this thesis proposes a grey Wolf algorithm to improve it at the same time.By using the improved grey Wolf algorithm,a new service discovery mechanism is proposed from two aspects of parameter optimization and service discovery strategy change.Gray Wolf algorithm can be divided into single objective gray Wolf algorithm and multi-objective gray Wolf algorithm,this thesis will improve single target gray Wolf algorithm applied to the service parameters optimization,combined with dynamic population combined with dynamic weighting of the mixed strategy for improvement,and using the improved single objective gray Wolf algorithm can quickly find the optimal solution through iteration characteristic,the reasonable optimization services registered discovery system related parameters,further optimize the performance of the system;In this thesis,the multi-objective grey Wolf algorithm used in large scale and dynamic stronger,service discovery scene under open computing environment,combined with double document and crossover and mutation operators of genetic algorithm,and choice of the three wolves group way of combining the strategy of improvement,using the improved multi-objective gray Wolf algorithm pheromones can store historical document controlled solution,and sensing network topology and the characteristics of the service resources change,combined with the storage service routing table search the hop,quickly locate the target service location,thus improving the efficiency of service discovery.Through the experimental results show that with the Gnutella and Random Walks are two classic unstructured P2 P network resource discovery algorithm,the algorithm in this thesis after the visit to around 30% of the registered node can obtain a high recall rate,and the other two kinds of classic algorithms due to the lack of historical information to guide the blind search,the recall rate is related to the number of nodes in the visited,basic geometric relations.After a lot of actual game plan tests,show that the proposed genetic algorithm solution in this work for the curling plan scheduling problem is effective.