自适应业务场景的数据库冷热数据识别算法  

Database cold and hot data identifying algorithm for transaction scenery adaption

作　　者：梁懿 陈又咏 李森[2] 王明 LIANG Yi;CHEN Youyong;LI Sen;WANG Ming(Fujian Yirong Information Technology Co.,Ltd.,Fuzhou 350003,China;School of Electronic and Information Engineering,Xi’an Jiaotong University,Xi’an 710049,China;State Grid Information&Telecommunication Group Co.,Ltd.,Beijing 102211,China)

机构地区：[1]福建亿榕信息技术有限公司,福建福州350003 [2]西安交通大学电子与信息学部,陕西西安710049 [3]国网信息通信产业集团有限公司,北京102211

出　　处：《现代电子技术》2022年第7期107-111,共5页Modern Electronics Technique

基　　金：国网信通产业集团两级协同研发企业级办公软件自主可控关键技术研究和应用示范项目资助。

摘　　要：作为数据库重要的底层技术,冷热数据识别旨在动态缓存热数据驻留内存、淘汰冷数据至外存,通过提高缓存命中率加快数据库访问响应速度。传统冷热数据识别方法如LRU、LFU主要针对时效性、频率性业务场景,当业务场景改变时,难以自适应新业务场景的冷热数据变化规律,因此缺乏灵活性。为此,提出HF-ARC算法,开展自适应业务场景的冷热数据识别。HF-ARC算法融合不同业务场景下的数据时效性与访问频率,实现了数据库冷热数据识别的自适应业务场景调优。基于Flash-DBSim系统仿真平台,分别采用局部、概率、线性三种访问模式模拟时效性、频率性、综合性业务场景。实验结果表明,HF-ARC算法根据不同业务场景冷热数据变化规律,动态调整冷热识别策略。相比非自适应识别算法,可将缓存命中率提高5%~15%;相比其他适应性算法,可将缓存命中率提高1.3%~2.5%。As an important underlying technology of the database,hot and cold data identification aims to dynamically cache hot data in memory,eliminate cold data to external storage,and speed up database access response speed by increasing the cache hit rate.The traditional cold and hot data identification methods,for example,LRU(least recently used)method and LFU(least frequently used)method,which are mainly target time-sensitive and frequent business scenarios.When the business scenario changes,it is difficult for them to adapt to the hot and cold data change laws of the new business scenario,so they lack flexibility.To this end,the HF-ARC(hot first-adaptive replacement cache)algorithm is proposed to identify hot and cold data in adaptive business scenarios.The HF-ARC algorithm integrates data timeliness and access frequency in different business scenarios,and realizes the adaptive business scenario optimization of database cold and hot data recognition.On the basis of the Flash-DBSim system simulation platform,the local access mode,probability access mode and linear access mode are used to respectively simulate timeliness,frequency and comprehensive business scenarios.The experimental results show that the HF-ARC algorithm can dynamically adjust the cold and hot identification strategy according to the changing laws of the cold and hot data in different business scenarios.In comparison with the non-adaptive recognition algorithms,the cache hit rate can be increased by about 5%~15%,while in comparison with the other adaptive algorithms,the cache hit rate can be increased by about 1.3%~2.5%.

关 键 词：数据库 冷热数据识别 自适应 系统存储 业务场景 缓存 局部性 

分 类 号：TN919.5-34[电子电信—通信与信息系统] TP333[电子电信—信息与通信工程]