高校云计算数据处理中虚拟机迁移与轮转模式研究

RESEARCH ON VIRTUAL MACHINE MIGRATION AND ROTATION MODE IN DATA PROCESSING OF COLLEGE CLOUD COMPUTING

针对高校云计算数据处理应用中的虚拟器迁移和数据处理任务调度问题,首先提出一种新颖且高效的数据依赖感知的虚拟机迁移方案(简称为DataAware)。在进行虚拟机迁移时,DataAware考虑了虚拟机之间的数据依赖性和服务器的剩余容量,以此减少迁移产生的网络流量。然后,为了进一步提高云计算中数据处理的性能,提出一种基于轮转模式的同步并行轮转(SPR)调度算法,以减少由数据汇聚而形成的通信瓶颈。最后分别采用仿真实验和理论分析评估所提出方法的性能。仿真实验结果表明,DataAware能够大大地减少网络流量,提高网络性能。通过推导出SPR策略的误差上界,说明采用SPR模式对数据处理任务进行调度能使数据处理算法收敛。

For the issues of virtual machine migration and data processing task scheduling during data processing in college cloud computing,this paper first proposes a novel and efficient Data Aware virtual machine migration algorithm(denoted by DataAware).While migrating the virtual machine,DataAware considers the data dependencies between virtual machines and the remaining capacity of the server to reduce the traffic caused by migration.Then,to improve the performance of data processing applications in cloud computing,we propose a synchronous parallel round robin(denoted by SPR)scheduling scheme to reduce the communication bottleneck caused by data aggregation.Finally,we perform simulation and theoretical analysis to evaluate our proposals.Simulation results show that DataAware can greatly reduce network traffic and improve network performance.We derive the upper bound of the error of SPR,which shows that the scheduling of data processing tasks by SPR mode can make the data processing algorithm converge.