高通量仿真中数据存储与访问技术研究  

Research on Technology of Data Storage and Access in High-throughput Simulation

作　　者：王子硕 翟岩龙[1] 陶文俊 杨浩 张晗 卿杜政[2] 

机构地区：[1]北京理工大学计算机学院,北京100081 [2]北京仿真中心航天系统仿真重点实验室,北京100854

出　　处：《系统仿真学报》2017年第9期2016-2024,共9页Journal of System Simulation

基　　金：国家自然科学基金(61602037)

摘　　要：针对大数据环境下仿真应用的海量数据处理需求,参照面向大数据高通量计算的架构和技术特点,提出并定义了高通量仿真的概念和参考架构。针对高通量仿真中数据访问瓶颈问题,设计了基于分布式内存文件系统的高通量仿真数据存储和访问体系,并扩展集成了非易失性存储器进一步提高数据访问的吞吐量。通过对典型仿真应用的实验验证,表明采用分布式内存文件系统和非易失性存储器的高通量仿真存储系统能够有效提升高通量仿真数据存储与访问性能。Aiming at the massive data processing requirements of simulation application in the big data environment, the concept and reference structure of high-throughput simulation were proposed and defined in accordance with the architecture and technical characteristics of high-throughput big data computing. For the problem of data access bottleneck in high-throughput simulation, a high-throughput simulation data storage and access system was designed based on distributed memory file system, and the non-volatile memory was integrated to improve the throughput of data access. The experimental results of typical simulation applications show that the high-throughput simulation storage system with distributed memory file system and non-volatile memory can effectively improve the data storage and access performance of high-throughput simulation.

关 键 词：大数据 高通量仿真 非易失性内存 分布式文件系统 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]