基于工作负载预测的虚拟机整合算法  

Workload Prediction-based Algorithm for Consolidation of Virtual Machines

作　　者：魏亮[1] 黄韬[1] 陈建亚[1] 刘韵洁[1] 

机构地区：[1]f北京邮电大学泛网无线通信教育部重点实验室,北京100876

出　　处：《电子与信息学报》2013年第6期1271-1276,共6页Journal of Electronics & Information Technology

基　　金：国家973计划项目(2011CB302901,2012CB315801);中央高校基本科研业务费专项资金(2011RC0118)资助课题

摘　　要：针对云计算环境下满足负载均衡、自动伸缩、绿色节能等需求时所面临的虚拟机(VM)迁移问题,该文设计一种面向云计算基础设施基于工作负载预测的整合调度算法。通过有机结合基于工作负载预测的主动控制技术和基于实际系统状态信息的被动控制技术,并采用指数平滑预测模型预测未来时刻的工作负载情况,提出虚拟机选择阶段最大未来工作负载优先和虚拟机安置阶段比较资源需求队列的虚拟机整合算法。仿真表明,该算法利用基于预测的资源整合方式减少了服务器使用量、虚拟机迁移次数和服务等级协议违例次数,有效提升了以数据中心为核心的云基础设施整体资源利用率。For issue of Virtual Machine （VM） migration in cloud computing environment when it comes to meeting the demands of load balancing, auto scaling, green energy-saving, etc. This paper design a scheduling algorithm that is cloud computing infrastructures oriented and workload prediction based. By organically integrating the active control technology based on workload prediction and the passive control technology based on status information of actual system, as well as with the exponential smoothing prediction model to predict the workload condition in future time, a VM consolidation algorithm is put forward which takes the maximum future workload as first in the VM selection stage and compares the resource demand queues in the VM placement stage. The simulation results show that the algorithm uses the prediction-based resource integration to reduce the number of servers and virtual machine migrations as well as service level agreement violations, effectively increasing the overall resource utilization of data center as the core of the cloud infrastructure.

关 键 词：云计算 基础设施即服务 工作负载预测 虚拟机整合 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]