基于负荷-容量模型的网络相继故障研究.pdfVIP

计算机研究 与发展 ’ ISSN 1000—1239／CN 11-1777门rP JournalofComputerResearchandDevelopment 49(12)：2529-2538，2O12 基于负荷一容量模型的网络相继故障研究 郭 迟 王丽娜。李 玉 周芙蓉。 (武汉大学卫星定位导航技术研究中心 武汉 430079) (武汉大学空天信息安全与可信计算教育部重点实验室 武汉 430072) 。(中国工商银行数据 中心 北京 100096) (guoehi@whu．edu．cn) Studyon NetworkCascadingFailuresBasedonLoad—CapacityM odel GuoChi ，W angLina ，LiYu ，andZhouFurong’。 (GlobalNavigationSatelliteSystem ResearchCenter，WuhanUniversity，Wuhan430079) (KeyLaboratoryofAerospaceInformationSecurityandTrustComputing (WuhanUniversity)，MinistryofEducation， V “口 430072) 。(DataCenterofIndustrialandCommercialBankofChinaLimited，Beijing100096) Abstract CascadingfailuresalwaysOCCUrincomputernetworksinwhich networktrafficisseverely impairedorhaltedtoorbetween largersectionsofthenetwork，caused byfailingordisconnected hardwareorsoftware． “Load—capacity’’modelsareusuallyusedforsolvingnetwork trafficproblems and exploring themechanismsof cascading failures． Centering on cascading failures in complex networks，thefollowingworkisdone．Firstly，theeffectofnetworktrafficloadoncascadingfailures isanalyzed． It indicates thatthecommunication activity among network nodespresents a self- organizedcriticalityphenomenon，andtheinfluenceonthenetworkrobustnessbroughtbythetraffic changeofthoseoriginalinactive nodesismuch greater than thatbroughtby thoseactive ones． Secondly，undertheconstraintsofeconomyandtechnology，acostfactorisintroducedtomodelthe relationship ofnetwork capacity and load，SO asto revealtheir constraintsrelationship．Finally， Centeringontheissueof “how toallocatethe1imitedredundantresourcestoanetworkwithspecific structurein orderto improve itsrobustness”，an evolutionary algorithm to search an optimized capacity—allocationstrategyisproposed，whichmakesthenetworkachieveoptimalrobustnesswiththe sameresources．Experimentsshow thatouralgorithm can findabetterresultofc