混合内存云架构的数据对象管理策略  

Data management strategy in mixed RAMCloud architecture

作　　者：岳珊 于炯[1,2] 鲁亮[2] 王跃飞[2] 蒲勇霖 YUE Shan;YU Jiong;LU Liang;WANG Yue-fei;PU Yong-lin(School of Software,Xinjiang University,Urumqi 830008,China;School of Information Science and Engineering,Xinjiang University,Urumqi 830046,China)

机构地区：[1]新疆大学软件学院,新疆乌鲁木齐830008 [2]新疆大学信息科学与工程学院,新疆乌鲁木齐830046

出　　处：《计算机工程与设计》2018年第2期381-386,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(61262088;61462079;61562086;61363083;61562078)

摘　　要：为有效解决内存云主要存储器件DRAM面临能耗损失大、数据可靠性差、断电易失等问题,提出使用非易失性存储器件PCM替换部分DRAM器件构成混合内存云架构。构建最优分配比模型,在系统总能耗一定的情况下,DRAM和PCM的分配比为4.847∶1时,系统的存储效用最高;通过数据对象预分类模型将数据对象按照读写倾向性分配到内存器件中,构建二室模型对数据对象的迁移和备份速率进行计算,确保系统的稳定性和可靠性。模拟算例结果表明,与相同能耗的内存云架构相比,该混合内存云架构提高了9.37%的存储能力,断电恢复延迟减少15.87%。To effectively solve DRAM as the main memory device of RAMCloud facing the problems of high energy consumption,poor data reliability,and easily disappearing data in the case of power outage,etc.,a design method using non-volatile PCM replacing part of DRAM devices constituting a mixed RAMCloud architecture was presented.Constructing the optimal allocation ratio model,the storage efficiency of the mixed RAMCloud system reached the highest when the total power consumption of the system was fixed and the distribution ratio of DRAM to PCM was 4.847∶1,the data objects were stored separately in different memory devices according to the read and write tendencies by the pre-classification model.The two-compartment model was constructed to calculate the migration and backup rate of the data objects so as to ensure the stability and reliability of the system.The simulation results show that the mixed RAMCloud architecture improves the memory capacity by 9.37%and reduces the power recovery delay by 15.87%compared with the original RAMCloud architecture

关 键 词：混合内存云 数据对象 最优分配比 数据预分类 二室模型 

分 类 号：TP393.02[自动化与计算机技术—计算机应用技术]