基于自适应控制方法的云计算服务器性能管理

Performance Management of the Cloud Server Based on Adaptive Control Approach

随着云计算技术的广泛使用,如何对采用虚拟化技术的云计算服务器的性能进行有效管理,是云计算研究的热点问题之一。论文提出了一种基于自适应控制理论的动态资源控制策略(DRC),该控制策略在保证服务级别协议的前提下,对运行在服务器上的各个虚拟机进行优化配置,使服务器的硬件资源得到最大化的利用。同时设计了一种新型的自适应线性二次高斯控制器,来应对具有Web应用所面对的动态负载。在基于Xen技术搭建的实验平台上,对服务器的性能在不同工作负载的情况下进行了测试,并与未采用DRC策略的服务器性能进行了对比。实验结果表明,在动态工作负载下,与为采用DRC策略的服务器相比,DRC控制策略能够有效保证不同Web应用的响应时间稳定在设定的参考值。

As the increasing interest on cloud computing, it is a hot topic in cloud computing that how to effectively manage the performance of the virtualized server. In this paper, a dynamical resource control solution（DRC） via adaptive control theory is proposed. It can efficiently optimize the resource allocation to co-located virtual machine（VM）, while meeting the service-level agreements（SLAs）. Meanwhile, for coping with the highly dynamical workload of Web services, a novel adaptive linear quadrie regulator （LQR） controller and a online estimator is designed. Our DRC solution on the Xerrbased testbed is implemented, and the performance of DRC solution in face of a variety of workload patterns is evaluated, compared with the virtualized server without DRC solution. The experiment results show that the virtualized server with DRC solution can effectively improve the performance of each VM, while guaranteeing the response time of each VM converge to the desired value