面向虚拟计算环境的负载均衡研究与设计

发布时间：2018-03-31 20:34

本文选题：虚拟计算环境　切入点：虚拟机迁移　出处：《中国舰船研究院》2012年硕士论文

【摘要】：虚拟化技术的出现，为提高当前数据中心等高性能计算系统的资源利用率提供了良好的解决方案。虚拟化数据中心中，由于应用程序的不确定性、服务器处理能力的差异性容易导致此类虚拟计算环境中服务器的负载失衡。负载均衡算法作为虚拟化多机环境下的关键技术，极大地影响着虚拟计算环境的整体性能。如果失去了负载均衡算法的支持，服务器负载不均衡，会出现单点服务器负载过重，而其它服务器闲置的问题。该单点服务器上的请求得不到及时处理，服务质量得不到保证，严重时会导致该服务器瘫痪。因此如何保持各个服务器的负载，使系统维持高效率运转就成了一个非常关键的问题。本文详细介绍目前虚拟机计算环境下负载均衡算法的相关技术，并分析传统集群环境中负载均衡策略与虚拟计算环境下负载均衡策略的不同。在Linjiu虚拟计算环境下对传统基于固定阈值的负载均衡算法进行初步设计现实，通过实验分析传统基于固定阈值的负载均衡算法的优缺点，提出面向虚拟计算环境的的负载均衡方法。具体分为两种应用模式，根据服务器平均负载的高低自动选择空闲时段策略和繁忙时段策略。在服务器平均负载空闲时段应用改进的基于固定阈值的负载均衡算法，用CPU预增量的形式能有效防止群聚效应。在服务器平均负载繁忙时段应用基于CPU负载平均的负载均衡算法，该算法在给定自由度内是相对自由的，只有当某些服务器负载差值过大时才触发发生迁移。两种模式的算法中都用到了CPU预测模型，通过预测负载的变化趋势可以避免呈显著下降趋势等情况下发生的不必要的迁移。通过仿真实验为CPU预测模型的应用选定具体参数，结果表明CPU预测模型的预测基本准确。两种不同应用模式下，不同负载均衡算法下的服务器状态不同，，结果显示本文设计的负载均衡方法能比较有效地利用两种算法各自的优点，使虚拟化数据中心的负载水平保持在均衡水平。
[Abstract]:The emergence of virtualization technology provides a good solution for improving the resource utilization of high performance computing systems such as current data centers. The difference of server processing ability easily leads to the load imbalance of the server in this kind of virtual computing environment. Load balancing algorithm is the key technology in virtualized multi-machine environment. It greatly affects the overall performance of the virtual computing environment. If the load balancing algorithm is lost, the server load will be unbalanced, and the single point server will be overloaded. The problem of idle other servers. The request on the single point server can not be processed in time, the quality of service can not be guaranteed, and the server will be paralyzed when serious. Therefore, how to maintain the load of each server, Keeping the system running efficiently is a key issue. This paper introduces the technology of load balancing algorithm in virtual machine computing environment in detail. The difference between load balancing strategy in traditional cluster environment and that in virtual computing environment is analyzed. In Linjiu virtual computing environment, the primary design of traditional load balancing algorithm based on fixed threshold is carried out. By analyzing the advantages and disadvantages of traditional load balancing algorithms based on fixed threshold, a load balancing method based on virtual computing environment is proposed. According to the average load of the server, the idle period strategy and the peak period strategy are automatically selected. The improved load balancing algorithm based on fixed threshold is applied in the server average load idle period. The clustering effect can be effectively prevented by using the form of CPU preincrement. The load balancing algorithm based on CPU load average is applied in the peak period of server average load, and the algorithm is relatively free within a given degree of freedom. The migration is triggered only when the load difference of some servers is too large. The CPU prediction model is used in both algorithms, and the unnecessary migration can be avoided by predicting the trend of load change. Through the simulation experiment, the specific parameters of CPU prediction model are selected. The results show that the prediction of CPU prediction model is basically accurate, and the server states under different load balancing algorithms are different under two different application modes. The results show that the load balancing method designed in this paper can effectively utilize the advantages of the two algorithms and keep the load level of the virtualized data center at the balance level.
【学位授予单位】：中国舰船研究院
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP338

【参考文献】