增强多目标遗传算法下的组合服务的容错能力  

Enhance fault-tolerant ability of Web services composition under multi-object genetic algorithm

作　　者：蔡美玲[1,2] 陈继锋[1,2] 

机构地区：[1]湖南涉外经济学院,长沙410205 [2]中南大学信息科学与工程学院,长沙410083

出　　处：《计算机工程与应用》2010年第32期1-5,28,共6页Computer Engineering and Applications

基　　金：国家自然科学基金(No.60970098;No.60803024;No.90715043);湖南省自然科学基金(No.10JJ6092);湖南省教育厅科学研究项目(No.08C517);湖南省第二批大学生研究性学习和创新性实验计划项目(湘教通[2009]320号;No.364)~~

摘　　要：利用多目标遗传算法得到一组满足约束条件的最优非劣方案集,用户根据实际需要从中选择满意的组合方案,同时,未被选择的流程作为备选方案以便发生意外时替代执行。提出了最优非劣方案的失效成功率和基本服务的相容性计算方法,选择可用性高且失效成功率较高的组合方案,以便执行发生意外时组合服务有较高的概率成功切换到备选方案继续执行,增强了组合服务执行的容错能力。实验结果说明了该方法的可行性和有效性。Pareto-based multi-object genetic algorithm for Web services selection achieves a set of pareto solutions which meet constraints,then the most approving solution is selected as the final decision project according to users＇requirements.At the same time,the solutions not chosen can be reserved as potential options in case of the failure while running.In this paper the calculating method of tolerability for pareto solutions and compatibility for basic service is proposed.The tolerabili-ty is provided as an evaluation index for selecting the final decision solution,where the higher availability and tolerability,the more probability it will be selected.Then more backup process can be switched in case of the running process occurring failure.Therefore fault-tolerant ability of Web services composition at run-time is enhanced.The experimental results indicate the feasibility and efficiency of this algorithm.

关 键 词：PARETO解集 容错 组合服务 多目标遗传算法 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]