基于FPGA的Memcached加速研究  

Study on Memcached Acceleration Based on FPGA

作　　者：韩玉浩 蒋越 HAN Yuhao;JIANG Yue(Department of Computer Science,National University of Defense Technology,Hunan 410073 China.)

机构地区：[1]国防科技大学计算机学院

出　　处：《集成电路应用》2020年第1期10-13,共4页Application of IC

基　　金：湖南省教育系统科技创新课题项目

摘　　要：作为一种高性能内存键值缓存系统,Memcached 对于提升 Web 服务质量至关重要并已广泛部署。因其基于 x86 架构的传统实现,在延迟和功耗等方面的问题日益显现,对其进行软件调优或硬件加速的尝试不断出现。研究基于 FPGA 的 Memcached 加速技术,与以往的研究相比,在系统结构和加速方案上,更注重可部署性。原型系统表明该系统的请求延迟降低至 3.2 μs,相比软件实现降低一个数量级,可实现 Memcached 请求 10 Gbps 线速处理。As a high-performance memory key-value caching system,Memcachedis very important for improving the quality of web service and has been widely deployed.Because of its traditional implementation based on x86 architecture,the problems of latency and power consumption are becoming more and more obvious,and the attempt of software tuning or hardware acceleration is constantly emerging.In this paper,we study the FPGA based Memcached acceleration technology.Compared with the previous research,the system structure and acceleration scheme are more flexible and deployable.The prototype system shows that the request delay of the system is reduced to 3.2 μs,which is an order of magnitude lower than that of the software implementation.The system achieves 10 Gbps line-speed processing.

关 键 词：集成电路设计 FPGA MEMCACHED 加速 低延迟 

分 类 号：TN42[电子电信—微电子学与固体电子学] TP333[自动化与计算机技术—计算机系统结构]