虚拟集群迁移优化绿色计算：能耗与效率提升策略

175 浏览量更新于2024-08-26 收藏 861KB PDF 举报

"基于虚拟集群的实时迁移在绿色计算中的虚拟资源整合是当前数据中心节能领域的重要研究课题。随着云计算的发展，数据中心承载的应用程序对能源消耗产生了巨大影响。为了实现绿色计算，关键在于灵活管理虚拟集群，从而降低整个数据中心的总能耗，并提高物理节点的运行效率。本文主要关注如何通过减少物理节点的平均功率开销时间以及提升其处理效率来达到节能减排的目标。具体而言，问题被转化为优化一个目标：在单位能量消耗下最大化数据处理量。作者提出了一个以自动电压控制器(AVC)为基础的迁移策略，该策略利用虚拟集群封装的并行任务完成时间作为指导，智能地调度虚拟机迁移。算法名为VCGM（Virtual Cluster Guided Migration），其核心原理是通过分析每个虚拟集群中任务的并发性和优先级，动态调整虚拟机的分配，使得那些完成时间较长的任务能在更高效、节能的物理节点上运行。这种迁移决策旨在平衡任务的执行时间和电力消耗，以期在满足服务质量和节能减排之间找到最佳平衡点。 VCGM算法通过对虚拟机进行实时监控和预测，能够在不中断服务的前提下，灵活调整虚拟集群的配置，从而实现实时的绿色计算资源优化。这不仅有助于降低数据中心的碳足迹，也有利于企业的运营成本控制和可持续发展。该研究提供了一种创新的解决方案，将虚拟化技术与节能策略相结合，旨在推动绿色计算的发展，为数据中心的能源管理带来实质性的改进。通过实施VCGM，可以期待在大规模虚拟化环境中实现更高的能源利用率，助力构建更加环保的信息技术基础设施。"

Virtual Resource Consolidation for Green Computing

Based on Virtual Cluster Live Migration

Xiaohui Wei, Yue Jin, Hongliang Li*, Xingwang Wang, Shoufeng Hu

College of Computer Science and Technology, Jilin University, Changchun 130012, China

Email: weixh@jlu.edu.cn; yuejinjy@yahoo.com; lihongliang@jlu.edu.cn;

wangxw2109@mails.jlu.edu.cn; shoufenghu@foxmail.com

Abstract—Data centers hosting applications consume

huge amounts of energy. Flexible management of virtual

clusters is an effective way to reduce the overall power

consumption of data centers. In order to save energy we

focus on decreasing the total power-on time of physical

nodes in data centers and increasing the runtime efficien-

cy of physical nodes. For a specific physical node how

long it is power-on is determined by the job with longest

completion time. In this work, we formulate the problem

into maximizing the amount of data which is processed

by unit energy. A VC-based migration approach is pro-

posed, which exploits the completion time of parallel jobs

encapsulated by virtual clusters to guide virtual machines

migrations. Algorithm VCGM is designed to construct the

final mapping state which is energy-efficiency. In VCGM,

all virtual clusters are grouped by their remaining time

and every virtual cluster group is “compacted” into few

physical nodes one by one. Then we adopt the low-cost

perfect matching in the bipartite matching problem to

obtain a migration plan. Algorithms are simulated and

evaluated to verify the effectiveness of our energy effi-

ciency scheme.

Index Terms—Virtual cluster migration, resource consolidation,

green computing

I. INTRODUCTION

Cloud computing has gained more and more attention

from industry and academic in recent years. It leverages

virtualization technology allowing customers to on-

demand provision resources based on a pay-as-you-go

utility model. Virtualization technologies have several

properties: encapsulation, isolation and hardware inde-

pendence, and so on. These properties make the way of

utilizing data centers’ computing resources in cloud

changed. Virtualization is the core technology of cloud

computing and views computing resources as a pool of

Manuscript received August 5, 2015; revised December 20, 2015.

This work was supported by the National Natural Science Founda-

tion of China (NSFC) (Grant No.61170004), IBM Shared University

Research (SUR) project (3D514BF41421), Specialized Research Fund

for the Doctoral Program of Higher Education (20130061110052), Key

Science and Technology Research Project of Science and Technology

Department of Jilin Province (20140204013GX), Special Fund for

Scientific Research in the Public Interest (SinoProbe09-01), and the

research supported by China Geological Survey project

(12120113006300).

*Corresponding author email: lihongliang@jlu.edu.cn.

doi:10.12720/jcm.v.n.p-p

unified resources, which reduces the complexity and

manageability of data centers.

Infrastructure providers maintain thousands of physical

computing nodes to meet the continuously increasing

requirement. Cloud computing brings convenience to

users and vendors, while we should not ignore the huge

energy consumption it brings too [1, 2]. The work [3]

estimates that a data center with 50,000 computing nodes

may consume more than one hundred million kwh/year

energy. This enormous energy consumption is equivalent

to the electricity consumption for a 100,000 population

urban in one year and it will generate heat and lead to

expensive cooling costs.

The reasons for this extremely high energy consump-

tion are: mass of computing resources, hardware power

inefficiency, and resources inefficient usage. Large num-

ber of underutilized servers has become a major problem

to cloud providers that needed to be figured out. As the

base power consumption is the dominant part of total

power consumption, no matter whether the node is busy

or idle the energy drawn by it is huge. How long a physi-

cal node is power-on depends on the job with longest

completion time. VM consolidation is an effective meth-

od to increase physical resources utilization and reduce

the power consumption. Live migration is a valid method

that migrate some VMs to other physical machines if

necessary for VM consolidation.

Virtual cluster (VC) [4, 5] is a group of VMs which are

used to complete a parallel job. Virtual cluster becomes

popular to run high performance computing workloads.

And its sub-tasks will run on several virtual machines.

The interaction operations between tasks are required for

synchronization. The parallel job is not completed until

all its tasks have been finished.

After a parallel job is assigned to a virtual cluster, the

required resources of physical nodes must be allocated to

all VMs belonging to this virtual cluster. The random

placement of VMs belonging to same virtual cluster may

cause some scattered resources fragment or some un-

derutilized physical nodes generated when the assigned

job finish. This is adverse to energy saving. At present,

the researches on virtual cluster mainly focus on its de-

ployment, but have little attention on its live migration.

Virtual cluster live migration concerns the overall compu-

ting environment’s migration. The impaction of migration

to parallel job performance should be considered.

Based on the mechanism Limevi [6] which supports

virtual cluster live migration in parallel computation of

下载后可阅读完整内容，剩余9页未读，立即下载

weixin_38670318

粉丝: 6
资源: 919

虚拟集群迁移优化绿色计算：能耗与效率提升策略

鸿鹄论坛_华为横纵虚拟化及集群技术培训

虚拟化的集群资源管理技术研究!.pdf

我有两个虚拟化集群在同一个vcenter上，想把一个集群中的虚机迁移到另一个集群上，应该怎么做

clickhouse集群数据迁移

elasticsearch集群数据迁移

多台服务器虚拟化集群

基于多虚拟控制点的无人机集群协同编队控制算法

mongodb6.0 分片集群 数据迁移

mongo3.6.6集群数据迁移

hadoop跨集群数据迁移

最新资源

mongodb6.0 分片集群数据迁移