基于效用的云计算容错策略和模型  

Utility-based fault tolerant strategy and model for cloud computing

作　　者：范贵生[1,2] 虞慧群[1] 陈丽琼[3] 刘冬梅[1] 

机构地区：[1]华东理工大学计算机科学与工程系,上海200237 [2]计算机软件新技术国家重点实验室,南京210093 [3]上海应用技术学院计算机科学与信息工程学院,上海200235

出　　处：《中国科学：信息科学》2014年第1期158-176,共19页Scientia Sinica(Informationis)

基　　金：国家自然科学基金(批准号:61173048;61300041);上海市教育委员会科研创新项目(批准号:12YZ166);高等学校博士学科点专项科研基金博导类资助课题(批准号:20130074110015);中央高校基本科研业务费专项基金(批准号:WH1314038);上海市科委重大项目(批准号:12510503800)资助项目

摘　　要：云计算近年来已成为一种被广泛接受的计算模式.随着云计算在商业、交通、卫生等领域应用的不断推进,云应用系统的可靠性问题引起了人们的特别关注.然而,云应用系统的结构和行为特征复杂,如何保障系统的可靠性是一项极具挑战性的课题.本文研究云计算的容错模型和策略,通过构建可扩展的云计算容错模型,以刻画云计算的运行机理、组件故障行为、云应用间合作和竞争特性.依据云计算的故障及资源服务特征,提出云计算的故障迁移和恢复方案.围绕容错涉及的时间和价格,依次计算云计算组件和云应用的效用,进一步分析各云应用的利益.通过求解模型的Nash均衡,以优化整体云计算的容错效用.最后,利用模型检查技术验证容错模型和容错处理的正确性.本文研究对于揭示云计算的结构和行为特征、建立云计算容错设计理论、提高云计算容错的效用具有理论意义和应用价值.Cloud computing has emerged as a widely accepted computing paradigm over the past few years. With the increasing application of cloud computing in vast areas such as industry,transportation and healthcare,the reliability of application systems draws special attention. However,due to the complexity of cloud computing system in structural and behavioral aspects,it is a great challenge to guarantee the reliability of applications in cloud environment. In this paper,we investigate the fault tolerant strategy and model for cloud computing. By constructing the extensible fault models of cloud computing systems,we can characterize the mechanisms of cloud computing,the fault behaviors of component,the cooperation and competition relations between cloud applications. Considering the fault and service resource of cloud computing,we solve the Nash equilibrium of the model and optimize the recovery program of cloud computing and propose the fault migration and restoration schemas for cloud computing. By considering the fault time and its cost,we compute the utility of the components of cloud computing and cloud application,and further analyze the benefts of cloud application. In order to optimize the fault tolerant efectiveness of cloud computing,we solve the Nash equilibrium of model. Model checking techniques are used to verify the correctness of fault diagnosis and recovery. This research is of both theoretical and practical signifcance in revealing structural and behavioral characteristics of cloud computing,and improving reliability of cloud computing systems.

关 键 词：云计算 容错 PETRI网 博弈 模型验证 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]