Virtual Machine Migration Strategy in Federated Cloud
Yanjue Xu, Sekiya Yuji · IEICE Technical Report; IEICE Tech. Rep. · 2010
Cloud computing and the as-a-service paradigm have gained a lot of interest recently. At the same time, virtualization technology has been shown to be an attractive path to increase overall system resource utilization by its powerful management mechanism such as isolating resources schedulers, suspend/resume, and virtual machine(VM) live migration. In this paper we present a new VM migration strategy using VM live migration technology in the federated cloud environment. This method will be used to detect the overloaded servers and initiate the migration to the optimized location in the cloud automatically, thus eliminating the hotspots and balancing the load not only CPU load, including memory and network utilization. According to the experimental result, our technique has been proven that it can detect and remove the hotspots efficiently and balance the load. K eyword Virtual Machine, Live Migration, Dynamic Resource Allocation, Load Balancing 1 . I n t rodu c t ion The pay-per-use model in the infrastructure as a service (IaaS) paradigm in cloud computing offers the ability to scale up compute and storage resources on demand. The Amazon Elastic Compute Cloud (Amazon EC2) is the best-known example of this paradigm of elastic capacity provisioning. By introducing virtualization to the clouds, users can be isolated with each other while sharing the same physical machine in the public cloud. Furthermore, Virtual Machine (VM) independence from hardware and support for heterogeneous software stacks has exempted cloud users from manual configuration. And the use of live VM migration technology has enabled more effective sharing of system resources across multiple physical severs. In the cloud environment, the workload of servers will fluctuate due to the incremental growth, time-of-day effects, and flash crowds. When a server is overloaded, Service Level Agreement (SLA) will start to degrade, which leads, for instance, that the response time of the request from the user will become longer. Therefore, how to allocate the virtual resources dynamically has become a widely concerned problem of cloud computing. Some researches about dynamic allocation of resources of grid computing have been working on allocating the processes to the CPU resources, rather than consider VM as an individual unit[1]. So it is difficult to directly apply then to the virtualized servers. The widely used cloud manager, such as OpenNebula[11], has solved the problem consider the automatic dynamic allocation while the workloads and demands of VMs fluctuate in real time using migration technology. The cloud user has to detect the hotspot and migrate it to a less loaded server manually. VMware DRS adopts the live VM migration technology to manage the operational cost of the cloud[6][10]. And since information of the mechanism of it is not available, and it can be only applied to the VMs running on the hypervisor from VMware. In order to address these problems, this paper proposes a dynamic virtual resources allocation mechanism using live VM migration. This approach automates tasks of monitoring the current load of servers and VMs, detecting the hotspots, deciding the best physical location of the busy VM, and initiating the migration. According to the experimental result, our technique has been proven that it can detect and remove the hotspots efficiently in the mostly under-loaded cloud, and balance the load in the mostly overloaded cloud. This paper is structured as follows: Section 2 presents the research background and system overview. In section 3, the results of experiments are showed and analyzed. And we will discuss about the current work and future work in section 4. インターネットコンファレンス2010 (IC2010) 2010年10月25日 10月26日