面向收益最大化的云资源自适应调度.pdfVIP

2014 年8 月第8 期 小型微型计算机系统 Journal of Chinese Computer Systems Vol. 35 No.8 2014 面向收益最大化的云资源自适应调度 冯国富，刘林源，唐明伟，韩冰青 (南京审计学院计算机系，南京 21∞29) E-mail: njufgf@gmail. com 摘 要:与网格、集群等传统计算模式相比，云计算的关键优势在于为用户提供了一种利用远程计算资源的实用商业模型.云 服务提供商把服务层指标转换为操作层指标，根据服务级别协议动态管理云计算资源.本文研究如何基于服务级别协议在不同 的客户之间动态分配云资源池资源，以获得最大收入.借助排队论模型，把资源分配描述为基于价格机制、服务请求到达率、服 务速率、可用资源等约束条件的数学极值问题，并通过拉格朗日乘子法给出了最优数学解.分别以合成数据和网络日志数据模 拟请求到达率，对理论结论进行了实验验证.实验结果表明本文算法有优于相关工作. 关键词:云计算;服务级别协议;资源分配;价格模型 中圄分类号:1凹93 文献标识码:A 文章编号:1侃lO-1220( 2014 )08-1717-05 Maximum Protit Oriented Adaptive Scheduling of Cloud RI臼ourc臼 FENG Guo-旬， LIU Lin-yu阻， TANG Ming-wei , HAN Bing-qing ( Department of Compuωr Science , Nanjing Audit University , Nanjing 21∞29 ， αIina) Abs钮8ct: A key advantage of cloud computing is 也就 it provides a 萨actical business model for customers to use remote reso田ωs， which distingt血hes cloud from traditional computing m叫els such as grid ∞m阴阳g and cluster ∞mputing. Cloud providers 缸ansform the service level metrics into 0严m出.g level me创cs ， and ∞ntrol the 萨)()led cloud reso田Ces adaptively among the differentiated cus- tomers based on Service Level Agr回ment (SLA). This pa萨~ addresses how ωmaximize the revenue 由rough SLA-based dyn缸nic resource allocation. We have fonnalized the resource all∞础。n problem with Queuing Theory , in which many constrains such as pri- cing mechanisms and available resources 缸e considered. The mathematical answer is alω 萨esented 也rough the method of Lagrange Multiplier. Two types of experiments , whose r吨ues臼 come 位。ms.归自.etic dataset and 国ced dataset res萨~tively ， have 以出en provided to validate 由e theoretical resu1ts. 四e experimental resu1ts show 也to町 algori也ms outperform relal时work. Key words: cloud compu出g; service level agreement; reso山田 d∞四on; pricing model