自适应的多虚拟机迁移调度算法

摘要

The multiple virtual machines (VMs) migration scheduling was studied in cloud data center.Combined with software defined networking (SDN), which decoupled the control and data planes, a migration framework based on SDN was proposed to improve the transfer speed of migration traffics.Aiming at minimizing the total migration time and service downtime, fine-grained factors (CPU, memory, bandwidth) were considered and a comprehensive selectivity factor based on Gauss formulation was proposed, which was the core of the proposed migration scheduling algorithm.To validate the efficiency and effectivity of the proposed method, extensive simulations were tested with the baselines in three scenarios.Compared with CQNCR algorithm, the results show that the proposed algorithm can decrease the total migration time and average VDC downtime by up to 30.7% and 23.1%, respectively.%针对云计算数据中心内部多虚拟机的迁移调度问题进行研究.结合SDN软件定义网络架构,利用其数据层和控制层分离、逻辑集中控制以及开放可编程接口的优点,设计一种基于SDN软件定义网络的多虚拟机迁移架构,提高虚拟机迁移流量的转发效率,阐明所提架构下,多虚拟机迁移的实现流程.综合考虑各种迁移影响因素,提出一个基于高斯形态的调度优先级因子,设计一种自适应的多虚拟机迁移调度算法AVMS(adaptive VMs migration scheduling).在3种不同规模的场景下进行仿真实验,实验结果表明,相比分组迁移算法CQNCR,AVMS在总体迁移时间和平均VDC宕机时间方面,分别降低最高达到30.7%和23.1%.