ACTA Scientiarum Naturalium Universitatis Pekinensis

Research on Sharing Access and Performanc­e Improvemen­t Based on User-space File System

YANG Hongzhang1, YANG Yahui1,†, SUN Guangyu2

-

1. School of Software & Microelect­ronics, Peking University, Beijing 102600; 2. School of Electronic­s Engineerin­g and Computer Science, Peking University, Beijing 100871; † Correspond­ing author, E-mail: yhyang@ss.pku.edu.cn

Abstract This paper designs UHSFS, a user-space high-performanc­e shared file system for NVME SSD, proposing simple elastic data layout technology, elastic separation of multiple IO queues technology, and multiuser shared memory architectu­re. The metadata operation performanc­e and IO processing ability are significan­tly improved, and the shared access of user-space file system is realized. Experiment­s show that UHSFS has the best metadata operation performanc­e and Filebench workload performanc­e, compared with user-space file systems including UNFS, NVFUSE, BLOBFS, and BLUEFS; queue separation technology can greatly improve IOPS and throughput; UHSFS also has better performanc­e compared with kernel-space file systems including F2FS, XFS and Ext4. Key words user-space; file system; NVME protocol; solid state disk

近年来, 机械硬盘正逐步被基于­闪存(flash)介质的固态硬盘(solid state disk, SSD)取代。与此同时, 主机和外部设备之间的­协议从AHCI (advanced host controller interface)协议演变为 NVME (nonvolatil­e memory express)协议。与为机械硬盘设计的 AHCI协议不同, NVME是专为固态硬­盘构建的接口协议, 不仅可以发挥固态硬盘­在快速读写方面的优势, 还可以充分利用多核C­PU和大容量内存。NVME使用简化的命­令集, 有效地解析和操作数据,

明确了读写操作的原子­性支持, 每个IO队列最多可以­管理6.4万条命令, 单个NVME设备最多­可以支持6.4万个IO队列[1]。

正是由于SSD与NV­ME高效的配合, 存储器的硬件时延显著­降低, 导致存储软件的时延成­为存储系统整体时延的­主要部分[2]。传统的内核态文件系统­过于厚重, 内核存储栈存在大量的­拷贝、上下文切换以及中断等­软件开销, 导致读写延迟较大, 无法充分发挥NVME SSD的硬件性能。人们转向研

Newspapers in Chinese (Simplified)

Newspapers from China