面向大数据的分布式缓存设计  

Design and Optimization of Distributed Caching System for Big Data

作　　者：董昭通 李小勇[1] DONG Zhao-tong;LI Xiao-yong(School of Cyber Security,Shanghai Jiao Tong University,Shanghai 200240,China)

机构地区：[1]上海交通大学网络空间安全学院

出　　处：《通信技术》2020年第1期114-119,共6页Communications Technology

基　　金：闵行区科技项目（No.2018MH331）~~

摘　　要：大数据平台的底层存储系统往往无法匹配上层计算应用的读写性能,而一个设计良好的分布式缓存系统将缩小CPU密集型应用和IO密集型应用之间不匹配的性能差距。设计的面向大数据应用的分布式缓存系统,在读写流程、I/O事件驱动并发模型及元数据模型等方面进行了合理设计与优化,并使用fio工具测试了顺序写、随机写、顺序读及随机读场景下的吞吐率与IOPS等性能指标,验证了该分布式缓存系统的高性能优势和应对高并发场景的扩展能力。The underlying storage systems of big data platforms often cannot match the read and write performance of upper-level computing applications.A well-designed distributed cache system will reduce the mismatched performance gap between CPU-intensive applications and IO-intensive applications.The distributed cache system for big data applications designed in this paper is reasonably designed and optimized in terms of read and write processes,I/O event-driven concurrency models,and metadata models.The fio tool is used to test performance indicators such as throughput and IOPS in sequential write,random write,sequential read,and random read scenarios.Finally,the high-performance advantages of the distributed cache system and the ability to scale in high-concurrency scenarios are verified.

关 键 词：分布式缓存 两级元数据模型 协程池 事件驱动并发模型 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]