欢迎来到天天文库
浏览记录
ID:51214382
大小:723.96 KB
页数:10页
时间:2020-03-21
《一种多租户云数据存储缓存管理机制.pdf》由会员上传分享,免费在线阅读,更多相关内容在工程资料-天天文库。
1、计算机研究与发展DOI:10.7544/issn1OOO一1239.2014.2O130789JournalofComputerResearchandDevelopment51(11):2528—2537,2O14一种多租户云数据存储缓存管理机制史玉良王捷(山东大学计算机科学与技术学院济南250101)(shiyuliang@sdu.edu.cn)AMulti_TenantMemoryManagementMechanismforCloudDataStorageShiYuliangandWangJie(SchoolofComputerScienceandTec
2、hnology,ShandongUniversity,-,inan250101)AbstractWiththepopu【arizationofcloudcomputing,softwareasaservice(SaaS)hasbecomeanimportantformofcloudcomputing.Memoryresourceownedbyeachdatanodeinthecloudisakeyresourcetoimprovedataaccessperformanceofmulti—tenantapplications.Therefore,memoryr
3、esourceshareandprovisioninghavereceivedalotofattentionfromSaaSproviders.Fortheserviceproviders,howtoreasonablyallocatememoryresourceineachdatanodeinordertoobtainhigherprofitswhileguaranteeingtenants’servicelevelagreement(SLA)hasbecomeachallenge.Addressingthechallenge,weproposeafram
4、eworkofmulti—tenantmemorymanagement(MTMM)forclouddatastorageandcorrespondingmemoryallocationmethod.Themethodtakesthemaximumprofitsserviceprovidercanobtainasatarget.Combinedwithtenants’SLAprofitmodels,theglobalmemoryallocationproblemisanalyzedandmodeledasanobjectiveoptimalproblem.C0
5、rrespondingtheprofitsserviceprovidercangetunderdifferentmemoryallocationstrategiesarepredictedthroughit.Consideringthecharacteristicsofmulti—tenantmemoryallocation,wesolvetheproblembyoptimizedgeneticalgorithminordertoimprovetheperformanceofthemethod.ComparedwiththetraditionalLRUmet
6、hodandmulti—tenantmemoryallocationmethodemployedinsinglenode,themechanismproposedinthispapercaneffectivelymanagememoryandprovidehigherprofitsforserviceproviders.Keywordssoftwareasaservice(SaaS);multi—tenancy;memorymanagement;clouddatastorage;servicelevelagreement(SLA)摘要随着云计算的普及,软件即
7、服务(softwareasaservice,SaaS)逐渐成为云计算的一种重要表现形式.云中数据节点的缓存是提高多租户应用数据访问性能的一种重要资源,缓存资源的共享和分配受到SaaS提供商的关注.对SaaS提供商而言,如何在多租户间有效地分配数据节点上的缓存资源,从而满足租户的服务水平协议(servicelevelagreement,SIA),获得更高的收益已成为一项挑战.为此,提出了多租户云数据存储缓存管理机制,以实现服务提供商收益最大化的目标,结合SIA收益模型,评估不同缓存策略下服务提供商获取的收益值,将全局缓存管理问题定义为目标优化问题,并结合缓存分
8、配特点,采用优化的遗传算法解决该问题.通过实验比较,
此文档下载收益归作者所有