Memcached的分析与改进  

Analyzing and improving memcached

作　　者：李首扬 杨仝[1] 代亚非[1] 黄亮 郑廉清 

机构地区：[1]北京大学计算机系 [2]94782部队65分队 [3]西京学院控制工程系

出　　处：《网络新媒体技术》2016年第6期38-42,共5页Network New Media Technology

基　　金：国家自然科学基金重点项目(编号:61232004);国家自然科学基金(61672061);国家"973"课题(编号2014CB340400);国家重点研发计划课题(2016YFB1000300);中国科学院声学研究所合作项目"分布式缓存客户端定制开发"

摘　　要：Memcached是一种高性能分布式内存对象缓存系统,使用十分广泛;其设计目的为缓存数据库内容以加速动态Web请求,但也广泛应用于其他高性能存储,并且成为了内存Key-Value数据库的经典代表。本文对Memcached的系统结构、处理流程进行了分析,包括网络处理、哈希表的查询与维护、内存的分配与管理、冷数据的置换算法等;同时以哈希表、内存管理这两个影响性能的关键点入手,结合了Cuckoo哈希与CLOCK置换算法两种更易并行化的的算法,对Memcached现有的结构提出了较大的改动,以期提高上述处理速度,实现整体性能的提升。Memcached, a widely used memory object caching system, is a typical case of in memory key - value storing or caching systems. This paper analyzed the architecture of memcached, including network processing, structure of the hash table, managing memory allocation, and the cache replacement algorithm. Focusing on two fields： the hash table structure, and the memory management politics, which have a significant impact on the performance and scalability of the system, this article introduces new algorithms, improving the performance of the parallelizing memory caching system. In this article, we combine Cuckoo hashing, a hashing scheme optimized for parallel accessing and scenes when read requests are greatly more than write requests; and a cache replacement algorithm, the clock algorithm, to archive higher processing rate, with an acceptable loss of cache hit rate.

关 键 词：内存缓存系统 MEMCACHED 哈希表 内存管理 缓存置换算法 

分 类 号：TP333.1[自动化与计算机技术—计算机系统结构]