期刊文献+

面向机群互连网络的智能网卡设计 被引量:4

The Implementation of Intelligent Adaptor Oriented Cluster Network
下载PDF
导出
摘要 在机群系统中,机群的互连网络性能对整个机群系统的性能有着至关重要的影响。机群系统要求互连网络具有高带宽、低延迟、高可靠等特性,因而通常需要研究专用的网络以满足这一要求。该文介绍了一种面向机群系统互连网络的基于i960VH的智能网卡的设计和实现方法,该网卡采用可编程逻辑器件EPLD实现网络接口设计,采用DMA方式提高网络传输带宽,采用全流水方式减小网络传输延迟,并通过将通讯协议下载到网卡上来执行以提高计算和通讯的重叠。实测表明,在不包括上层协议的情况下,网卡到网卡的传输延迟为3us,节点到节点数据传送带宽可达500Mbit/s。 In cluster system,the performance of the interconnection network impacts the performance of system critically.Cluster system calls for such features as high bandwidth,low latency,high reliability,so need develop special network to meet the demands.This paper presents a kind of implementation of intelligents adaptor for cluster network based i960VH,the adaptor implements the NIC design by EPLD,takes advantage of DMA transfer to increase the bandwidth,exerts full pipelined transfer to decrease the latency,and downloads the transfer protocol to improve the overlap between computing and communication.Testing in real environment shows that in case of excluding the upper protocol,the adaptor-to-adaptor latency is3us,the node-to-node bandwidth is over500Mbps by using DMA transfer mode.
出处 《计算机工程与应用》 CSCD 北大核心 2003年第28期148-150,190,共4页 Computer Engineering and Applications
基金 国家863高技术研究发展计划项目:面向网格的高性能计算机-曙光4000的研究(编号:2002AA104410)
关键词 智能网卡 机群 互连网络 i960VH Intelligent adaptor,Cluster,Network,i960VH
  • 相关文献

参考文献5

  • 1安学军,祝明发,高文学,吴冬冬.虫洞路由交换及其缓冲区设计[J].微电子学与计算机,2002,19(9):15-18. 被引量:4
  • 2Scalable Coherent Interface[S].IEEE Std 1596--1992,IEEE CS Press, 1993.
  • 3N J Boden et al.Myrinet:A Gigabit Per-Second Local-Area Network [J].IEEE Micro, 1995 ; 15( 1 ) :29-36.
  • 4Fabrizio Petrini,Wuchun Feng et al.The Quadrics Network:High- performance Clustering Technology[J].IEEE Micro,2002:46-57.
  • 5Intel Corporation.i960(r)VH Processor Developer's Manual.1998-10.

二级参考文献1

共引文献3

同被引文献17

  • 1王佰玲,方滨兴,云晓春.零拷贝报文捕获平台的研究与实现[J].计算机学报,2005,28(1):46-52. 被引量:67
  • 2刘勇,杨卫兵,陈曜,孙凝晖.智能网卡驱动程序的性能评价[J].计算机工程,2005,31(14):42-44. 被引量:4
  • 3Xilinx Corp.RocketIO^TM Transceiver User Guide[Z].2002-10.
  • 4SAWICKI T.Technology strategist,corporate technology group,Intel corporation.server network I/O acceleration:fundamental to the data center of the future[J].Technology@Intel Magazine,2004:1-7.
  • 5EARLS A.Integrating TCP offload engines (TOEs)[EB/OL].2002-05.[2006-06-08].http://ajilonconsulting.ajilon.com/pdf/IntegratingTCPOffloadEngines0,pdf.
  • 6BRADNER S,McQUAID J.RFC2544[EB/OL].1999-03.[2006-06-08].http://www.ietf.org/rfc/rfc2544.txt.
  • 7BRADNER S.RFC1242[EB/OL].1997-07.[2006-06-08].http://www.faqs.org/rfcs/rfc1242.html.
  • 8Tanabe N,Hamada Y,Nakajo H,et al.A low latency high bandwidth network interface prototype for PC cluster∥Proceedings of the International Workshop on Innovative Architecture for Future Generation High-Performance Processors and Systems.Big Island,2002:87
  • 9Tanabe N,Yamamoto J,Nishi H,et al.On-the-fly sending:a low latency high bandwidth message transfer mechanism∥Proceedings of the International Symposium on Parallel Architectures.Dallas,2000:186
  • 10JEDEC Standard.PC1600 and PC2100 DDR SDRAM Unbuffered DIMM Design Specification.Revision 1.1.[2000-12-15].http:∥download.micron.com/pdf/toolbox/184ddrunbrev1.1.pdf

引证文献4

二级引证文献5

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