基于会话共享的负载均衡方案研究

发布时间：2018-05-18 18:47

本文选题：集群 + 负载均衡　；参考：《山东科技大学》2017年硕士论文

【摘要】：伴随着网络的飞速发展,用户访问量急剧增长,对Web服务器的性能和可靠性提出了更大的挑战,使用服务器集群技术和负载均衡技术能有效的解决大量web并发访问给后端服务器造成压力的问题。其中负载均衡技术是服务器集群系统的核心,它关系到后端服务器能否保持高效可靠稳定的工作状态。客户端访问web时,由于HTTP协议是无状态的连接协议,Web网站会使用会话技术来记录客户端与服务器端的交互信息。当处于负载均衡的集群环境中时,通常会使用会话保持技术实现会话的一致性。这时,负载均衡算法和会话保持方法都会对负载均衡集群的性能及负载平衡性有影响。本文首先介绍了负载均衡技术的基本概念特点和算法,分析了常用的会话保持技术,对常用的会话保持技术的优缺点进行了分析,最后采用独立服务器管理会话数据,使会话数据以全局可见的方式来保持会话一致性,由于会话服务器将被各个节点频繁的访问成为系统中的热点之一,所以会话服务器采用冗余备份的方式。给出了系统的总体设计方案,从需求分析入手描述了整个会话的操作流程。采用Nginx服务器作为负载均衡服务器,分析了 Nginx的源代码,采用一种动态的负载均衡算法,将节点服务器的剩余负载和性能比值转化为权重,选出权重前三的服务器,然后在被选出的服务器中随机选出一台分配客户端请求,随机分配时,被选中的概率与服务器的权重成正比,从而增强了集群的性能,提高了效率。最后,利用LoadRunner性能测试工具对原有加权轮询策略和本文所采用动态策略进行性能测试,通过实验证明,本文所采用的这种动态策略更好的实现了负载均衡。
[Abstract]:With the rapid development of the network, the number of users' access has increased dramatically, which poses a greater challenge to the performance and reliability of the Web server. Using server cluster technology and load balancing technology can effectively solve the pressure problem of a large number of Web concurrent access to the back end server. At the core, it is related to whether the back end server can maintain a high efficient, reliable and stable state of work. When the client visits the web, the HTTP protocol is a stateless connection protocol, and the Web site uses session technology to record the interaction information between the client and the server side. In this paper, the load balancing algorithm and session retention method will affect the performance and load balance of the load balancing cluster. Firstly, the basic concepts and algorithms of load balancing technology are introduced, and the common session retention technology is analyzed, and the advantages and disadvantages of the commonly used session retention technology are divided. Finally, the session data is managed by the independent server to keep the session consistency in a global visible way. Because the session server will be frequently visited by each node as one of the hotspots in the system, the session server uses redundant backup. The overall design scheme of the system is given and the requirements are divided. The Nginx server is used as the load balancing server. The source code of the Nginx is analyzed. A dynamic load balancing algorithm is used to transform the ratio of the residual load and the performance ratio of the node server to the weight, and the first three servers are selected and then selected randomly in the selected server. When assigning a client request, the selected probability is proportional to the weight of the server, thus enhancing the performance of the cluster and improving the efficiency. Finally, the LoadRunner performance testing tool is used to test the original weighted polling strategy and the dynamic strategy adopted in this paper. This dynamic strategy makes it better to achieve load balancing.
【学位授予单位】：山东科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.09

【参考文献】