Research on compatibility of high-performance network stack

Huiyou JIANG; Junfeng LI; Dan LI

doi:10.11959/j.issn.1000-0801.2019096

您当前的位置：

首页 >

文章列表页 >

Research on compatibility of high-performance network stack

Topics:intelligent communication technologies and applications | 更新时间：2024-06-05

- Research on compatibility of high-performance network stack
- Telecommunications Science Vol. 35, Issue 5, Pages: 25-31(2019)
- 作者机构：
- 作者简介：
- 基金信息：
- DOI：10.11959/j.issn.1000-0801.2019096
  CLC： TP393
- Published Online：2019-05，
  
  Published：20 May 2019
- 稿件说明：
移动端阅览
Huiyou JIANG, Junfeng LI, Dan LI. Research on compatibility of high-performance network stack[J]. Telecommunications science, 2019, 35(5): 25-31.
DOI：

Huiyou JIANG, Junfeng LI, Dan LI. Research on compatibility of high-performance network stack[J]. Telecommunications science, 2019, 35(5): 25-31. DOI： 10.11959/j.issn.1000-0801.2019096.

摘要

人工智能系统的海量数据处理需要很高的网络吞吐来支撑，网络协议栈作为上层人工智能应用和底层高速硬件网卡之间的桥梁，对整个系统的网络性能起到关键性作用。通过对主流网络协议栈兼容性进行详细对比分析，发现目前各类协议栈兼容性普遍较低，导致人工智能应用移植困难等问题。提出了高性能用户态网络协议栈的高兼容性设计构想，使各类传统应用无需修改源码即可获得网络性能的提升，讨论了高性能协议栈的后续发展方向。

Abstract

The massive data processing in artificial intelligence system requires the support of high-throughput network.As a bridge between the upper artificial intelligence applications and the underlying high-speed hardware network adapter

the network stack plays a crucial role in the performance of the whole network system.Through detailed analysis and comparison of the compatibility of mainstream network stacks

it is found that the compatibility of various network stacks is generally low

which leads to difficulties in transplantation of artificial intelligence applications.The idea of achieving both high-compatibility and high-performance of user-level network stack were proposed.In this way

legacy applications could directly attain the improvement of performance without modification of the source code.Path for the subsequent development of network stack was lit up.

关键词

Keywords

references

ZHANG H , ZHENG Z , XU S , et al . Poseidon:an efficient communication architecture for distributed deep learning on GPU clusters [C ] // Annual Technical Conference,July 12-14,2017,Santa Clara,CA,USA.[S.l.:s.n] . 2017 : 181 - 193 .

KERRISKK M . The so_reuseport Socket option [EB ] . 2013 .

BARON J . Epoll:add EPOLLEXLUSIVE flag [EB ] . 2015 .

Epoll:add EPOLLEXLUSIVE flag [EB ] . 2017 .

PESTEREV A , STRAUSS J , ZELDOVICH N , et al . Improving network connection locality on multicore systems [C ] // The 7th ACM European Conference on Computer Systems,Apr 10-13,2012,Bem,Switzerland . New York:ACM Press , 2012 : 337 - 350 .

HAN S , MARSHALL S , CHUN B G , et al . MegaPipe:a new programming interface for scalable network I/O [C ] // The 10th USENIX Conference on Operating Systems Design and Implementation,Oct 8-10,2012,Hollywood,CA,USA . New York:ACM Press , 2012 : 135 - 148 .

LIN X , CHEN Y , LI X , et al . Scalable kernel TCP design and implementation for short-lived connections [J ] . ACM SIGPLAN Notices , 2016 , 51 ( 4 ): 339 - 352 .

THEKKATH C A , NGUYEN T D , MOY E , et al . Implementing network protocols at user level [J ] . IEEE/ACM Transactions on Networking , 1993 , 1 ( 5 ): 554 - 565 .

BASU A , . A user-level network interface for parallel and distributed computing [C ] // The 15th ACM Symposium on Operating Systems Principles (SOSP),Dec 3-6,1995,Copper Mountain,Colorado,USA . New York:ACM Press , 1995 : 40 - 53 .

EDWARDS A , MUIR S . Experiences implementing a high performance TCP in user-space [J ] . ACM SIGCOMM Computer Communication Review , 1995 , 25 ( 4 ): 196 - 205 .

MOGUL J , RASHID R , ACCETTA M . The packer filter:an efficient mechanism for user-level network code [C ] // The Eleventh ACM Symposium on Operating Systems Principles,Nov 8-11,1987,Austin,USA . New York:ACM Press , 1987 : 39 - 51 .

Intel . Data plane development kit project page [EB ] . 2012 .

RIZZO L , . Netmap:a novel framework for fast packet I/O [C ] // 21st USENIX Security Symposium,Aug 8-10,2012,Bellevue,WA,USA . New York:ACM Press , 2012 : 101 - 112 .

NTOP . Multi-10 Gbit RX/TX packet processing from hosts and virtual machines [EB ] . 2016 .

JEONG E Y , WOOD S , JANSHED M , et al . mTCP:a highly scalable user-level TCP stack for multicore systems [C ] // USENIX Conference on Netwoked Systems Design ＆Implentation,Apr 2-4,2014,Seattle,WA,USA . New York:ACM Press , 2014 : 489 - 502 .

BELAY A , PREKAS G , KLIMOVIC A , et al . IX:a protected dataplane operating system for high throughput and low latency [C ] // USENIX Conference on Operating Systems Design ＆Implementation,Oct 6-8,2014,Broomfield,CO,USA . New York:ACM Press , 2014 : 49 - 65 .

PREKAS G , KOGIAS M , BUGION E . ZygOS:achieving low tail latency for microsecond-scale networked tasks [C ] // The 26th Symposium on Operating Systems Principles,Oct 28-31,2017,Shanghai,China . New York:ACM Press , 2017 : 325 - 341 .

PETER S , LI J , ZHANG I , et al . Arrakis:the operating system is the control plane [J ] . ACM Transactions on Computer Systems (TOCS) , 2016 , 33 ( 4 ):11.

HONDA M , HUICI F , RAICIU C , et al . Rekindling network protocol innovation with user-level stacks [J ] . ACM SIGCOMM Computer Communication Review , 2014 , 44 ( 2 ): 52 - 58 .

YASUKATA K , HONDA M , SANTRY D , et al . StackMap:low-latency networking with the OSstack and dedicated NICs [C ] // 2016 USENIX Annual Technical Conference,Jun 22-24,2016,Denver,CO,USA . New York:ACM Press , 2016 : 43 - 56 .

POPE S , RIDDOCH D . Introduction to OpenOnload—building application transparency and protocol conformance into application acceleration middleware [R ] . 2011 .

PKELSEY . A library version of FreeBSD’s TCP/IP stack plus extras [EB ] . 2016 .

TECENTE . F-stack [EB ] . 2018 .

SCYLLADB . High performance server-side application framework [EB ] . 2014 .

YULINTAO . Nginx with mTCP and DPDK [EB ] . 2016 .

SHALEV L , SATRAN J , BOROCIK E , et al . IsoStack-highly efficient network processing on dedicated cores [C ] // USENIX Annual Technical Conference,June 23-25,2010,Boston,MA,USA . New York:ACM Press , 2010 :5.

Views

589

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Research and design of distributed high-performance network reptiles based on cloud platform

Compatibility study between IMT and inter-satellite service systems in 25.25～27.5 GHz frequency band

Related Author

Enming SHI

Xiaojun XIAO

Yu LU

Zhaojun QIAN

Tan WANG

Long KANG

Bo LI

Related Institution

State Radio Monitoring Center

North China Electric Power University

Beijing University of Posts and Telecommunications

⁰