超级计算机网络引导技术研究与分析

doi:10.11772/j.issn.1001-9081.2018122605

计算机应用 ›› 2019, Vol. 39 ›› Issue (6): 1577-1582.DOI: 10.11772/j.issn.1001-9081.2018122605

• 2018全国高性能计算学术年会(HPC China 2018)论文 • 上一篇下一篇

超级计算机网络引导技术研究与分析

龚道永, 宋长明, 刘沙, 漆锋滨

江南计算技术研究所, 江苏无锡 214083

收稿日期:2018-12-12 修回日期:2019-02-28 发布日期:2019-06-17 出版日期:2019-06-10
通讯作者: 龚道永
作者简介:龚道永(1973-),男,安徽颍上人,高级工程师,硕士,主要研究方向:并行操作系统、系统容错、功耗管理;宋长明(1981-),男,黑龙江齐齐哈尔人,助理研究员,硕士,主要研究方向:并行操作系统、系统容错;刘沙(1984-),男,江苏盐城人,工程师,硕士,主要研究方向:并行操作系统、人工智能;漆锋滨(1966-),男,江西南昌人,研究员,博士,CCF会士,主要研究方向:计算机体系结构、编译器。
基金资助:
国家重点研发计划项目（2017YFB0202004）。

Research and analysis of supercomputer network boot technology

GONG Daoyong, SONG Changming, LIU Sha, QI Fengbin

Jiangnan Institute of Computing Technology, Wuxi Jiangsu 214083, China

Received:2018-12-12 Revised:2019-02-28 Online:2019-06-17 Published:2019-06-10
Supported by:
This work is partially supported by the National Key Research and Development Program of China (2017YFB0202004).

摘要/Abstract

摘要： 针对超级计算机系统中网络引导时间开销大的问题，提出网络引导分布算法是影响网络引导性能的主要因素之一，是优化网络引导性能的主要方向的观点。首先，分析了影响大规模网络引导性能的主要因素；其次，结合一种典型超级计算机系统，分析了超节点循环分布算法（SCDA）和插件循环分布算法（BCDA）的网络引导数据流拓扑结构；最后，量化分析了这两种算法对各个网络路径段的压力和可获得的网络性能，发现BCDA性能是SCDA性能的1~20倍。通过理论分析和模型推导发现，在计算节点和引导服务器之间使用更细粒度的映射算法可以在引导部分资源时使用尽量多的引导服务器，减少对局部网络资源的过早竞争，提升网络引导性能。

关键词: 超级计算机, 插件板, 超节点, 网络引导开销, 全路径最小带宽, 网络引导分布算法

Abstract: Since the network booting time overhead is high in supercomputer system, the idea that the network boot distribution algorithm is one of the main factors affecting the network boot performance and the main direction of optimizing network boot performance was proposed. Firstly, the main factors affecting large-scale network boot performance were analyzed. Secondly, combined with a typical supercomputer system, the network boot data flow topologies of Supernode Cyclic Distribution Algorithm (SCDA) and Board Cyclic Distribution Algorithm (BCDA) were analyzed. Finally, the pressure of above two algorithms on each network path branch and the available network performance were quantitatively analyzed. It can be seen that the bandwidth performance of BCDA is 1-20 times of that of SCDA. Theoretical analysis and model deduction show that the finer-grained mapping algorithm between compute nodes and boot servers can make as many boot servers as possible be used while boot some resources, reducing the premature competition for partial network resources and improving network boot performance.

Key words: supercomputer, board, supernode, network boot overhead, full path minimum bandwidth, network boot distribution algorithm

中图分类号:

TP338.6

龚道永, 宋长明, 刘沙, 漆锋滨. 超级计算机网络引导技术研究与分析[J]. 计算机应用, 2019, 39(6): 1577-1582.

GONG Daoyong, SONG Changming, LIU Sha, QI Fengbin. Research and analysis of supercomputer network boot technology[J]. Journal of Computer Applications, 2019, 39(6): 1577-1582.

参考文献

[1] 刘沙,龚道永.超级计算机低功耗技术研究[J].高性能计算技术,2014,227(2):60-67.(LIU S, GONG D Y. Research of lowpower techniques for HPC[J]. Journal of High Performance Computing Technology, 2014, 227(3):60-67.)
[2] 毛樟根,雷鸣.一种MPP系统开工的设计与实现[J].高性能计算技术,2011,210(3):30-34.(MAO Z G, LEI M. A kind of design and practice of MPP system management[J]. Journal of High Performance Computing Technology, 2011, 210(3):30-34.)
[3] Red Hat Company. The Red Hat kickstart installation method[EB/OL].[2018-10-20]. http://fedoraproject.org/wiki/Anaconda/Kickstart.
[4] Intel Corporation. Preboot Execution Environment (PXE) specification version 2.1[EB/OL].[2018-10-20]. http://www.pix.net/software/pxeboot/archive/pxespec.pdf.
[5] RFC 1350. The TFTP protocol (revision 2)[S/OL].[2018-10-20]. http://www.rfc-editor.org/info/rfc1350.
[6] RFC 2131. Dynamic host configuration protocol[S/OL].[2018-10-20]. http://www.rfc-editor.org/info/rfc2131.
[7] RFC 3530. Network File System (NFS) version 4 protocol[S/OL].[2018-10-20]. http://www.rfc-editor.org/info/rfc3530.
[8] 刘青昆,阴元友,郑晓薇,等.基于KickStart的Linux网络安装新模型[J].计算机工程,2008,34(21):103-105.(LIU Q K, YIN Y Y, ZHENG X W, et al. New model to Linux network installation based on Kickstart[J]. Computer Engineering, 2008, 34(21):103-105.)
[9] 顾梦非,徐炜民.一个基于Linux的集群部署方案[J].计算机应用与软件,2008,25(1):102-104.(GU M F, XU W M. A deployment system for clusters based on Linux[J]. Computer Applications and Software, 2008, 25(1):102-104.)
[10] 李怀刚,邱建新.网络安装Linux的技术原理分析及实现[J].计算机应用与软件,2006,23(9):109-111.(LI H G, QIU J X. The realization and principle of Linux's install through network[J]. Computer Applications and Software, 2006, 23(9):109-111.)
[11] 王春耕,付金辉,朱建涛.大规模集群系统中基于Image的系统安装[J].高性能计算技术,2007,187(4):44-48.(WANG C G, FU J H, ZHU J T. Installation of operating system based on image in large scale cluster system[J]. Journal of High Performance Computing Technology, 2007, 187(4):44-48.)
[12] 周虎廷,罗茂盛,周达民.集群计算机系统部署分析与研究[J].高性能计算技术,2013,223(4):56-59.(ZHOU H T, LUO M S, ZHOU D M. Analysis and research of the deployment of cluster computer system[J]. Journal of High Performance Computing Technology, 2013, 223(4):56-59.)
[13] HENDRIKS E A, MINNICH R G, CHOI S, et al. Pink:a 1024-node single-system image Linux cluster[C]//Proceedings of the 2004 International Conference on High Performance Computing and Grid in Asia Pacific Region. Piscataway, NJ:IEEE, 2004:454-461.
[14] NGUYEN T L, LEBRE A. Virtual machine boot time model[C]//Proceedings of the 25th Euromicro International Conference on Parallel, Distributed and Network-based Processing. Piscataway, NJ:IEEE, 2017:430-437.
[15] 董小社,孙发龙,李纪云,等.基于映像的集群部署系统设计与实现[J].计算机工程,2005,31(24):132-134.(DONG X S, SUN F L, LI J Y, et al. Design and implementation of image based cluster deployment system[J]. Computer Engineering, 2005, 31(24):132-134.)

超级计算机网络引导技术研究与分析

Research and analysis of supercomputer network boot technology

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 3

编辑推荐

Metrics

[1]	田宝华蒋句平李宝峰张晓明屈婉霞. 基于统一资源管理的超级计算机系统节能方案[J]. 计算机应用, 2012, 32(03): 835-838.
[2]	谭义红陈治平李学勇林亚平. 基于k-完美差异图的超节点拓扑结构构造[J]. 计算机应用, 2011, 31(08): 2021-2024.
[3]	陈军刘佳琦陈志刚. 超节点覆盖网中针对free-rider的用户评价机制[J]. 计算机应用, 2008, 28(2): 360-362,.