期刊文献+

Redis集群系统的优化研究与实现 被引量:7

Research and Implementation of Optimization of Redis Cluster System
下载PDF
导出
摘要 现代互联网应用数据量大,并发高,对响应速率等要求极高。Redis作为高性能的分布式非关系型数据库,在互联网应用中表现极佳。现今Redis集群方案Redis Cluster有很多优越性,越来越多的项目都期望从Redis迁移到Redis cluster上,但如果直接迁移,工作量和复杂性极高。因此在研究对比当今流行的Redis集群系统架构设计和系统优化方案后,通过自研代理中间件Vc_redisproxy在代理中间件中利用主从Reactor线程模型对原生Redis内部单一的NIO多路复用模型进行优化,使用MainReactor处理连接请求,然后将连接交由SubReactor,由SubReactor来处理读写请求。然后使用线程池,通过增加IO线程数来充分发挥多核CPU的优势,同时在代理中维护一张数据路由映射表,保证IO效率最大化。基于Redis cluster使用Java语言优化并实现了一种Redis集群系统。最后使用Redis-benchMark工具对系统进行压力测试。结果表明,单个Vc_redisproxy性能相对于Codis代理性能提升了17%,同时采用优化后的系统框架后,部署两个Vc_redisproxy代理性能相比部署两个Codis代理性能提升了15%。 Nowadays,in Internet applications,there is a large amount of data,high concurrency,and extremely high requirements on response rate.As a high-performance distributed non-relational database,Redis performs very well in Internet applications.The Redis cluster solution now has many advantages.More and more projects are expected to migrate from Redis to Redis cluster,but if they are directly migrated,the workload and complexity are extremely high.Therefore,after researching and comparing today’s popular Redis cluster system architecture design and system optimization scheme,the self-developed proxy middleware-Vc_redisproxy uses the master-slave Reactor thread model in the proxy middleware to the single NIO multiplexing model in the native Redis To optimize,use MainReactor to process the connection request,and then hand over the connection to SubReactor,which will process the read and write requests.Then use the thread pool to give full play to the advantages of multi-core CPUs by increasing the number of IO threads,while main taining a routing map table in the agent to ensure maximum IO efficiency.A Redis cluster system is optimized and implemented based on Redis cluster using Java language.Finally,the Redis-benchMark tool is used to stress test the system.The results show that the performance of a single Vc_redisproxy is improved by 17%relative to the performance of the Codis proxy.At the same time,after using the optimized system framework,the performance of deploying two Vc_redisproxy proxies compared to the performance of deploying two Codis proxies increase by 15%.
作者 伍华锋 王小刚 候劲 唐小林 WU Huafeng;WANG Xiaogang;HOU Jin;TANG Xiaolin(School of Automation and Information Engineering,Sichuan University of Science&Engineering,Zigong 643000,China)
出处 《四川轻化工大学学报(自然科学版)》 CAS 2020年第6期39-45,共7页 Journal of Sichuan University of Science & Engineering(Natural Science Edition)
基金 国家自然科学基金项目(61902268) 四川省科技计划项目(2018JY0197,20ZDYF0919) 四川省教育厅项目(18ZA0357) 德阳市开放式校市合作技术研发项目(2018CKJSD017) 人工智能四川省重点实验室项目(2017RZJ02) 四川理工学院人才引进项目(2017RCL12)。
关键词 Redis集群 Reactor模式 路由映射 自研代理中间件 Redis cluster Reactor mode route mapping self-developed agent middleware
  • 相关文献

参考文献19

二级参考文献49

  • 1Do K D. Formation control of multiple elliptical agents with limited sensing ranges[J].{H}AUTOMATICA,2012,(7):1330-1338. 被引量:1
  • 2Manathara J G,Ghose D. Rendezvous of multiple UAVs with collision avoidance using consensus[J].{H}JOURNAL OF AEROSPACE ENGINEERING,2012,(4):480-489. 被引量:1
  • 3Zhou Z,Fang H,Hong Y. Distributed estimation for moving target based on state-consensus strategy[J].{H}IEEE Transactions on Automatic Control,2013,(8):2096-2101. 被引量:1
  • 4Boyd S,Ghosh A,Prabhakar B. Randomized gossip algorithms[J].{H}IEEE Transactions on Information Theory,2006,(6):2508-2530. 被引量:1
  • 5Dimakis A G,Sarwate A D,Wainwright M J. Geographic gossip:efficient averaging for sensor networks[J].{H}IEEE Transactions on Signal Processing,2008,(3):1205-1216. 被引量:1
  • 6Tuncer C A,Mehmet E Y,Anand D S. Broadcast gossip algorithms for consensus[J].{H}IEEE Transactions on Signal Processing,2009,(7):2748-2761. 被引量:1
  • 7Ustebay D,Oreshkin B N,Coates M J. Greedy gossip with eavesdropping[J].{H}IEEE Transactions on Signal Processing,2010,(7):3765-3776. 被引量:1
  • 8Kar S,Moura J M F. Gossip and distributed Kalman filtering:weak consensus under weak delectability[J].{H}IEEE Transactions on Signal Processing,2011,(4):1766-1784. 被引量:1
  • 9Cai K,Ishii H. Average consensus on general strongly connected digraphs[J].{H}AUTOMATICA,2012,(11):2750-2761. 被引量:1
  • 10Lavaei J,Murray R M. Quantized consensus by means of gossip algorithm[J].{H}IEEE Transactions on Automatic Control,2012,(1):19-32. 被引量:1

共引文献28

同被引文献35

引证文献7

二级引证文献16

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部