数据中心计算资源节能算法研究

发布时间：2018-01-09 05:35

本文关键词：数据中心计算资源节能算法研究　出处：《电子科技大学》2013年硕士论文　论文类型：学位论文

【摘要】：目前，数据中心普遍存在高耗能、资源浪费严重等问题。如何高效整合计算资源和降低能耗成本是数据中心研究的热点。由于虚拟化技术能够实现对资源的高效利用和简单管理，所以数据中心广泛采用基于虚拟机迁移的节能算法对资源进行分配和调度，实现服务器上的资源负载均衡和降低数据中心能耗的目的。然而，现有针对数据中心节能的算法大多没有考虑服务器之间虚拟机迁移的能耗开销，过度的依赖虚拟机迁移可以实现资源的平均分配，但有可能达不到降低系统能耗的目的。因此，，研究虚拟机迁移过程中的能耗开销和设计合理的节能调度算法对于数据中心能耗管理具有十分重要的意义。具体来讲，本文研究的主要内容和创新包括以下几点： 1.建立数据中心服务器能耗模型。本文对影响数据中心服务器能耗的各项因素（CPU利用率，内存利用率，磁盘读写情况等）进行逐一分析，并结合实验数据，建立服务器能耗模型； 2.建立虚拟机迁移能耗模型。本文对虚拟机迁移过程中的迁移性能、虚拟机性能损耗以及虚拟机迁移对服务器能耗的影响三个方面进行研究，建立虚拟机迁移能耗模型。 3.节能算法设计。本文在建立服务器能耗模型和虚拟机迁移能耗模型的基础上，设计了离线负载跨度最大节能算法和在线迁移节能算法。前者主要通过考虑负载的区间跨度，实现对虚拟机资源的合理分配；后者主要通过虚拟机迁移，以较少数量的服务器满足请求分配，达到降低数据中心总能耗的目的。 4.虚拟机迁移能耗实验验证。通过对虚拟机迁移过程中服务器能耗数据的采集和分析，验证了虚拟机迁移过程中存在能耗开销。实验结果显示，CPU利用率、VM内存大小和网络带宽对虚拟机迁移过程中的服务器能耗存在较大影响。 5.节能算法对比。通过与负载均衡节能算法、在线DRR节能算法、在线EAM节能算法和MBFD节能算法进行对比，实验结果表明，本文设计的离线负载跨度最大节能算法和在线迁移节能算法分别在数据中心总能耗、服务器开启总时间、服务器开启总数量和请求的拒绝次数等方面具有明显的优势。
[Abstract]:At present, high energy consumption is prevalent in data centers. How to efficiently integrate computing resources and reduce the cost of energy consumption is the focus of data center research. Because virtualization technology can achieve efficient use of resources and simple management. Therefore, energy saving algorithm based on virtual machine migration is widely used in data center to allocate and schedule resources to achieve resource load balance and reduce data center energy consumption on the server. Most of the existing algorithms for energy saving in data centers do not take into account the energy cost of virtual machine migration between servers. Excessive reliance on virtual machine migration can achieve equal allocation of resources. But it may not achieve the goal of reducing system energy consumption. Therefore, it is very important for data center energy management to study the energy consumption cost in the virtual machine migration process and to design a reasonable energy saving scheduling algorithm. The main contents and innovations of this paper include the following: 1. Establish the data center server energy consumption model. This paper analyzes the factors that affect the data center server energy consumption, such as CPU utilization, memory utilization, disk reading and writing. Based on the experimental data, the energy consumption model of the server is established. 2. Build the model of virtual machine migration energy consumption. This paper studies the migration performance, virtual machine performance loss and the impact of virtual machine migration on server energy consumption. Build the model of virtual machine migration energy consumption. 3. Energy-saving algorithm design. This paper establishes the model of server energy consumption and virtual machine migration energy consumption model. The maximum energy saving algorithm of off-line load span and the energy saving algorithm of online migration are designed. The former realizes the rational allocation of virtual machine resources by considering the interval span of load. The latter mainly migrates through virtual machines to satisfy the request allocation with fewer servers to reduce the total energy consumption of the data center. 4. Virtual machine migration energy consumption experimental verification. Through the collection and analysis of server energy consumption data during virtual machine migration process, verify the virtual machine migration process energy consumption overhead. The experimental results show. CPU utilization and VM memory size and network bandwidth have great influence on server energy consumption during virtual machine migration. 5.Compared with load balancing energy-saving algorithm, online DRR energy-saving algorithm, on-line EAM energy-saving algorithm and MBFD energy-saving algorithm, the experimental results show that. The maximum energy saving algorithm of off-line load span and the energy saving algorithm of online migration are designed in this paper, respectively in the data center total energy consumption, the total time to open the server. The total number of server openings and the number of requests rejected have obvious advantages.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP308

【参考文献】