基于Alluxio远程场景下缓存策略的优化  

Optimization of cache strategy based on Alluxio remote scenario

作　　者：董文菁 温东新[1] 张展[1] Dong Wenjing;Wen Dongxin;Zhang Zhan(School of Computer Science&Technology,Harbin Institute of Technology,Harbin 150001,China)

机构地区：[1]哈尔滨工业大学计算机科学与技术学院,哈尔滨150001

出　　处：《计算机应用研究》2018年第10期3025-3028,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(61370085)

摘　　要：当今时代数据呈现出指数级增长效应,更多的组织采用多数据中心和分布式来存储数据,Alluxio作为以内存为中心的虚拟分布式存储系统,整合了底层大数据生态系统。在Alluxio与底层存储结合的远程场景中,由于网络的延迟,使得I/O速度成为影响对外服务的重要因素之一。针对以上研究提出一种基于Alluxio远程场景下的缓存策略CPR,利用存储系统中数据块之间的关联性指导数据预取与替换,采用分组思想提高关联规则的利用率,启用后台线程实时更新规则集,并通过仿真实验验证策略的有效性。仿真结果表明,CPR策略指导下的I/O性能要优于Alluxio现有的缓存策略和一些基于数据块间关联规则的缓存策略。At present,the data show an exponential growth effect,more organizations use multiple data centers and distributed way to store data.Alluxio integrates the underlying data ecosystem as a memory-centric virtual distributed storage system.In the remote scene combined with Alluxio and the underlying storage,due to the delay of the network,the I/O speed becomes one of the important factor that affect the external service.This paper developed a cache strategy CPR based on the Alluxio remote scenario,used the association between data blocks in the storage system to guide the data prefetching and replacement,used the idea of grouping to improve the utilization rate of association rules,enabled the background thread to update the rule set in real time,and verified the validity of the strategy by simulation experiment.The simulation results show that the I/O performance under the CPR strategy is superior to Alluxio’s existing cache strategy and some caching policies based on inter-data association rules.

关 键 词：数据块 关联规则 Alluxio 数据预取 数据替换 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]