期刊文献+

异构非易失性内存卷模式实现与应用 被引量:1

IMPLEMENTATION AND APPLICATION OF HETEROGENEOUS NONVOLATILE MEMORY VOLUME MODEL
下载PDF
导出
摘要 大数据背景下,以计算为中心的系统架构逐渐转向以数据为中心,内存-磁盘模式已不能满足需求。基于非易失性内存设备的优秀特性,设计了非易失性内存卷模式改变传统内存-磁盘模式。热数据频繁访问导致严重访存冲突,降低系统性能。基于此,在非易失性内存卷模式基础上设计了一对多地址映射,通过将热数据备份放置在不同内存通道的非易失性内存上,多核环境中,对于同一块数据的多个访问请求分配到不同数据备份上,实现并行访问,降低系统访问延迟。通过实验在双通道内存架构中实现一对二的映射方式,结果表明这样的设计使得系统性能提升了7.25%。 Under the background of big data, the computing-centric system architecture has gradually shifted to the data-centric, memory-disk model has been unable to meet the demand. Based on the outstanding features of non-volatile memory devices, a non-volatile memory volume mode was designed to change the traditional memory-disk mode. Frequent access to hot data leads to serious memory access conflicts, which reduces system performance. Based on this, a one-to-many address mapping is designed based on the non-volatile memory volume mode. By placing the hot data backup on the non-volatile memory of different memory channels, under the multi-core environment access requests are assigned to different data backups, enabling parallel access and reducing system access latency. Experiments in the dual-channel memory architecture to achieve one-to-two mapping, the results show that this design makes the system performance increased by 7.25%.
作者 钱璐 李弋 吴毅坚 赵文耘 Qian Lu,Li Yi,Wu Yijian,Zhao Wenyun(1.School of Software, Fudan University, Shanghai 201203,China;2.Shanghai Key Laboratory of Data Science, Fudan University, Shanghai 200433, Chin)
出处 《计算机应用与软件》 北大核心 2018年第8期43-49,共7页 Computer Applications and Software
基金 国家高技术研究发展计划项目(2015AA015303)
关键词 非易失性内存卷 load/store访问方式 热数据 访存性能 Nonvolatile memory volume Load/store access method Hot data Performance of memory access
  • 相关文献

参考文献1

二级参考文献50

  • 1Qureshi M K, Gurumurthi S, Rajendran B. Phase change memory: from devices to systems[M]. San Rafael: Morgan & Claypool Publisher, 2011.
  • 2Fusion IO. The fusion-io difference[EB/OL]. [2015-05-06]. http://www.fusionio.com/load/- media-/lqaz4e/docsLibrary/FIO SSD Differentiator_ Overview.pdf.
  • 3Yang J, Minturn D B, Hady F. When poll is better than interrupt[C]//Conferenee on File and Storage Technologies (FAST). San Jose, CA, USA: USENIX, 2012: 25-32.
  • 4Nellans D, Zappe M, Axboe J, et al. Ptrim 0+ exists 0: Exposing new FTL primitives to applications[C]//The 2nd Annual Non-Volatile Memory Workshop (NVMW). La Jolla, CA, USA: UCSD, 2011: 17-17.
  • 5Prabhakaran V, Rodeheffer T L, Zhou L. Transactional flash[C]//Proceedings of the 8th USENIX Conference on Operating Systems Design and Implementation (OSDI). Berkeley, CA, USA: USEN1X, 2008: 147-160.
  • 6Ouyang X, Nellans D, Wipfel R, et al. Beyond block I/O: Rethinking traditional storage primitives[C]//Proceedings of the 17th IEEE International Symposium on High Performance Computer Architecture (HPCA). San Antonio, Texas, USA: IEEE, 2011: 301-311.
  • 7Lu Y, Shu J, Gun J, et al. LightTx: A lightweight transactional design in flash-based SSDs to support flexible transactions[C]//Proeeedings of the IEEE 31st International Conference on Computer Design (ICCD). Asheville, North Carolina, USA: IEEE, 2013:115-122.
  • 8Swanson S, Caulfield A M. Refactor, reduce, recycle: Restructuring the I/O stack for the future of storage[J]. Computer. 2013, 46(8): 52-59.
  • 9陆游游.闪存文件系统关键技术研究[D].北京:清华大学,2015.
  • 10Hewlett Packard Enterprise. StoreServ7450[EB/OL]. [2015-05-01]. http://www.hp.condhpinhdnewsroom/press.kits/2014/HPDiscover2014/3PAR.

共引文献26

同被引文献5

引证文献1

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部