内存计算环境下基于索引结构的内存优化策略  

An Optimized Storing Strategy Based on Index Structure for In-Memory Computing

作　　者：英昌甜 王维庆[2,3] 于炯 卞琛[2,4] 国冰磊[2] 祁雷[2] 

机构地区：[1]新疆大学电气工程学科博士后科研流动站,新疆乌鲁木齐830046 [2]新疆大学软件学院,新疆乌鲁木齐830046 [3]新疆大学电气工程学院,新疆乌鲁木齐830046 [4]乌鲁木齐职业大学信息工程学院,新疆乌鲁木齐830002

出　　处：《新疆大学学报（自然科学版）》2018年第1期13-21,共9页Journal of Xinjiang University(Natural Science Edition)

基　　金：新疆维吾尔自治区自然科学基金项目(2017D01A20)

摘　　要：由于内存计算能够较好的满足在线数据密集型应用的需求,近年来受到了研究者的广泛关注.内存云存储数据时使用哈希结构来提高写入和恢复效率,然而该结构会降低系统读性能,同时增加系统清理回收内存的开销.为了解决这个问题,提出一种基于索引压缩存储的内存优化策略.在存储时,将内存划分为两部分,哈希存储和排序存储.在系统繁忙时,对于实时写入和更新的数据存储时采用占用空间较多、插入效率较高的哈希存储;在系统空闲时段时,利用基于索引压缩的排序存储算法,将哈希存储转换为占用内存空间较少、查找效率较高的排序存储.实验结果表明,同未进行优化的Tachyon单一哈希结构存储策略相比,该策略能够很好地均衡系统的写入和读取访问的效率.Recently, in-memory computing has been attracting increasing attention because it can satisfy the requirements of the online data intensive application. Hash structure could improve the write efficiency of the storage system, at the same time limit the access efficiency. In order to solve the storage problem,an optimized strategy based on index compact storage was proposed to optimize the storage of data index.Firstly, memory of severs were divided into two parts, hash store and sorted store. When the system is busy, the new writing and updating data are store in the hash store which has high insertion efficiency but occupy much space. When the system is idle, by using sorted store algorithm, convert the hash store into the sorted store which has high access efficiency and occupy little space. The experiment demonstrates that comparing with non-optimization Tachyon, the proposed strategy is the proper trade-off between write and access efficiency.

关 键 词：内存计算 存储策略 哈希结构 排序存储 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]