Research And Application Of Xen-based Resource Dynamic Optimization Algorithm For Virtual Machines

In recent years, virtualization technology is widely used in various filed, such as resource service integration, system security, distributed computing, emerging cloud computing etc. Virtualization technology attracts the interests of researchers for its advantages in masking the heterogeneous characteristics of the hardware resource, improving the use efficiency of resource, enhancing system flexibility and scalability, reducing the hardware investment and so on. However, resource utilization may be extraordinary low or high in the runtime of each virtual machine. Besides, each physical node may be load imbalance in the virtualization platform. Based on the deep analysis of the Xen architecture, this paper researches the re-allocation of the CPU and memory resource. Further, this paper analyzes the physical node loading balance with the help of dynamic migration technology of virtual machines.Firstly, this paper deeply studies the Xen virtualization technology and proposes a fine-grained resource adjustment algorithm, to efficiently cope with the inefficient use of resources and unreasonable resources allocation, caused by the static allocation of resources. Based on the real-time statistical analysis of the utilization of the resource in each single physical node, this paper dynamically recycles resource from the virtual machine with low utilization and increases the amount of resource allocation for the virtual machine with high utilization. Extensive experiments show that our algorithm not only resolves the problem of low utilization of resources and unreasonable resource allocation effectively, but also avoids the unnecessary migration of virtual machines.Secondly, this paper proposes a VM-based coarse-grained migration scheduling algorithm to deal with the loading imbalance between the physical nodes. The coarse-grained migration algorithm mainly includes the selection of virtual machines to be migrated, the mapping relationship of migration destination and the migration's execution. The selection of virtual machines to be migrated prefers to those with low migration costs and great impact on physical machine's load. Based on the standard particle swarm algorithm, this paper redesigns available functions to meet various resources balanced and improves the calculation formula of the distance between particles. The improved particle swarm optimization algorithm can effectively detect the mapping relationship between virtual machine and target host. And migration execution algorithm reduces the total migration time by finding the optimal execution order of migration in parallel with the help of the migration conflict detection.Thirdly, a framework for dynamic optimization of the virtual machine resource is designed and implemented, which proposes the hotspot detection strategy and eliminates hot spots combined with, the above-mentioned two kinds of dynamic resource optimization algorithm. According to our analysis, this framework not only improves virtualization platform resource efficiency and user service quality but also reduces the management and maintenance costs. Further, it avoids the conflicts between the two particle size optimization algorithms.Finally, extensive experiments of the proposed dynamic resource optimization algorithm are executed to verify their efficiency. The experimental results show that the dynamic resource optimization algorithm can effectively solve the problem of the unreasonable resource allocation and load balancing and other related issues.