物流信息平台高并发技术研究

发布时间：2018-03-31 06:10

本文选题：高并发　切入点：数据库　出处：《北京邮电大学》2014年硕士论文

【摘要】：随着互联网的飞速普及,基于web的物流信息平台更加广泛的被使用。随之而来的是系统并发访问量的增长,系统的并发性能成为衡量物流信息平台的关键指标,同时也成为制约物流信息平台发展的瓶颈之一。本文分析了物流信息平台的主要子模块与实现技术,针对不同的模块进行了研究并提出了优化方案。在数据库方面,基于Mysql,针对Mysql事务实现方式的不足,提出了基于悲观锁和乐观锁编程模型。分析了表与索引的原理,针对表结构设计,索引设计提出了一般性的设计原则。在水平拓展方面,提出了垂直拆分与水平拆分技术方案。在缓存系统方面,比较了本地缓存与分布式缓存,提出了细粒度的缓存方案,同时结合物流信息平台中常见的业务逻辑设计了具有通用性的关联数据缓存方案。在Web服务器优化方面,研究了衡量web服务器性能的关键指标。针对JavaEE框架,研究了JVM内存回收机制,提出了JVM设置与垃圾回收器选择的一般性原则。并以tomcat为例,分析了线程数与I/0模型对web服务器性能的影响。此外对JAVA的NIO进行了研究,最后提出从线程数与I/O方式两方面对JavaEE web服务器进行优化的方案。在分布式集群方面,比较了不同的负载均衡实现方法。提出了三种集群间用户状态同步的方案。最后,基于上述理论对两个实际的高并发系统的进行了优化,取得了良好的优化结果,证明了上述理论的有效性。
[Abstract]:With the rapid popularization of the Internet, the logistics information platform based on web has been more widely used. With the increase of concurrent access to the system, the concurrent performance of the system has become the key index to measure the logistics information platform. At the same time, it has become one of the bottlenecks restricting the development of logistics information platform. In this paper, the main sub-modules and implementation techniques of the logistics information platform are analyzed, the different modules are studied and the optimization scheme is put forward. In the aspect of database, aiming at the deficiency of Mysql transaction implementation, the programming model based on pessimistic lock and optimistic lock is put forward based on MySQL. The principle of table and index is analyzed, and the design of table structure is given. The general design principles of index design are put forward. In the aspect of horizontal extension, vertical split and horizontal split technology are proposed. In the aspect of cache system, this paper compares the local cache with the distributed cache, proposes a fine-grained cache scheme, and designs a universal associated data cache scheme combined with the common business logic in the logistics information platform. In the aspect of Web server optimization, the key indexes to measure the performance of web server are studied. Aiming at the JavaEE framework, the mechanism of JVM memory recovery is studied, and the general principles of JVM setting and garbage collector selection are put forward. The tomcat is taken as an example. The influence of thread number and I / O model on the performance of web server is analyzed. In addition, the NIO of JAVA is studied, and the scheme of optimizing JavaEE web server from two aspects of thread number and I / O mode is put forward. In the aspect of distributed cluster, different load balancing methods are compared, and three schemes of user state synchronization among clusters are proposed. Finally, based on the above theory, two practical high concurrent systems are optimized, and good optimization results are obtained, which proves the validity of the above theory.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.05

【参考文献】