分布式缓存可用性相关问题研究

发布时间：2018-04-11 05:01

本文选题：分布式缓存 + 负载均衡　；参考：《华中科技大学》2013年硕士论文

【摘要】：在人类对网络需求越来越大的今天，简单的依赖客户端-服务器-数据库模式构建网络应用已经无法支撑日益增长的网络负载。网络缓存是提升Web系统整体性能理想的解决方案之一，，分布式缓存系统的相关研究从上世纪90年代开始，至今许多研究成果已经应用于各个大型网络应用中。分布式缓存具有良好的可扩展性、高可用性和结构对用户透明的特点。过于集中或严重倾斜的负载会使分布式缓存系统出现“热点”给单个节点造成巨大压力。均衡分组策略(Balanced Partition Scheme, BPS)在一致性哈希算法的基础上进行改进，解决了由于一致性哈希算法的随机性导致存储空间分配不合理的问题。该策略将哈希函数的值域空间平均分为若干子空间，并且根据节点的性能分配存储空间，保证处理能力强的节点能够分配到较多的请求负载，使系统在动态条件下保持负载均衡。基于系统可靠性的缓存数据冗余策略，根据可靠性理论将分布式缓存系统抽象成并联系统，在保证系统整体的可靠性基础上测算系统数据需备份数，并将缓存数据冗余存储在请求的转发节点中以保证缓存系统在动态情况下的命中率。本文利用分布式缓存系统对一个基于B/S结构开发的供水网信息管理系统进行性能优化，该系统为用户提供查询漏失率的功能。在分析系统用户请求习惯的基础上，增加数据预处理机制，将预处理结果写入分布式缓存系统中供客户端直接查询，并且利用Memcached分布式缓存系统对预处理机制进行仿真实验。仿真实验结果证明均衡分组策略具有较好的负载均衡效果，在忽略次要因素的情况下缓存冗余策略可以提高缓存系统的命中率，利用分布式缓存对信息管理系统进行优化可以有效的减少用户请求数据的时间，提高了服务器处理请求的效率。
[Abstract]:Nowadays, with the increasing demand for network, simply relying on client-server-database schema to construct network application can no longer support the increasing network load.Network caching is one of the ideal solutions to improve the overall performance of Web system. The research on distributed cache system has been applied to various large-scale network applications since the 1990s.Distributed cache has good scalability, high availability and transparent structure to users.Over-concentrated or heavily skewed loads can cause a "hot spot" in distributed caching systems, which can put a great deal of pressure on a single node.The balanced Partition Scheme is improved on the basis of the consistent hash algorithm, which solves the problem of unreasonable storage space allocation due to the randomness of the consistent hash algorithm.In this strategy, the range space of the hash function is divided into several subspaces, and storage space is allocated according to the performance of the node, so that the node with strong processing power can be allocated more request load.Make the system maintain load balance under dynamic conditions.Based on the buffer data redundancy strategy of system reliability, the distributed cache system is abstracted into parallel system according to the reliability theory, and the backup number of system data is calculated on the basis of guaranteeing the reliability of the system as a whole.The cache data redundancy is stored in the request forwarding node to ensure the cache system hit ratio dynamically.In this paper, a distributed cache system is used to optimize the performance of a water supply network information management system based on B / S structure. The system provides users with the function of query loss rate.On the basis of analyzing the user request habits of the system, the data preprocessing mechanism is added, and the preprocessing results are written to the distributed cache system for the client to inquire directly. The simulation experiment of the preprocessing mechanism is carried out by using the Memcached distributed cache system.The simulation results show that the group balancing strategy has a better load balancing effect, and the cache redundancy strategy can improve the hit rate of the cache system, while ignoring the secondary factors.Using distributed cache to optimize the information management system can effectively reduce the time of user request data and improve the efficiency of server processing requests.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP333

【参考文献】