面向收益最大化的云资源自适应调度

Maximum Profit Oriented Adaptive Scheduling of Cloud Resources

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.

A key advantage of cloud computing is that it provides a practical business model for customers to use remote resources, which distinguishes cloud from traditional computing models such as grid computing and cluster computing. Cloud providers transform the service level metrics into operating level metrics, and control the pooled cloud resources adaptively among the differentiated cus- tomers based on Service Level Agreement （ SLA ）. This paper addresses how to maximize the revenue through SLA-based dynamic resource allocation. We have formalized the resource allocation problem with Queuing Theory, in which many constrains such as pri- cing mechanisms and available resources are considered. The mathematical answer is also presented through the method of Lagrange Multiplier. Two types of experiments, whose requests come from synthetic dataset and traced dataset respectively, have been provided to validate the theoretical results. The experimental results show that our algorithms outperform related work.