支持低成本快速局部重构的快速Benes网络

秦梦远; 刘宏伟; 郝沁汾

doi:10.11959/j.issn.1000-0801.2026003

您当前的位置：

首页 >

文章列表页 >

支持低成本快速局部重构的快速Benes网络

研究与开发 | 更新时间：2026-02-07

- 支持低成本快速局部重构的快速Benes网络
- Fast Benes network that supports fast low-cost partial route reconfiguration
- 电信科学 2026年42卷第1期页码：86-104
- 作者机构：
  
  1.中国科学院计算技术研究所，北京 100095
  2.中国科学院大学，北京 101408
- 作者简介：
  
  [ "秦梦远（1994- ），男，中国科学院计算技术研究所博士生，主要研究方向为新型计算机系统结构、Benes网络、可重构光互连系统等。" ]
  [ "刘宏伟（1984- ），男，博士，中国科学院计算技术研究所高级工程师，主要研究方向为高性能处理器设计、异构计算、硬件安全与加密芯片、软件定义芯片等。" ]
  [ "郝沁汾（1969- ），男，中国科学院计算技术研究所正高级工程师、博士生导师，主要研究方向为高性能计算机、高端SMP服务器、CPU等。" ]
- 基金信息：
  
  国家重点研发计划项目(2022YFB4401501);江苏省重点研发计划项目(BE2023006-4)
- DOI：10.11959/j.issn.1000-0801.2026003
  中图分类号： TN401;TP302.2
- 收稿：2025-03-17，
  
  修回：2025-07-10，
  
  录用：2025-07-24，
  
  纸质出版：2026-01-20
- 稿件说明：
移动端阅览
秦梦远,刘宏伟,郝沁汾.支持低成本快速局部重构的快速Benes网络[J].电信科学,2026,42(01):86-104.

Qin Mengyuan,Liu Hongwei,Hao Qinfen.Fast Benes network that supports fast low-cost partial route reconfiguration[J].Telecommunications Science,2026,42(01):86-104.
秦梦远,刘宏伟,郝沁汾.支持低成本快速局部重构的快速Benes网络[J].电信科学,2026,42(01):86-104. DOI： 10.11959/j.issn.1000-0801.2026003.

Qin Mengyuan,Liu Hongwei,Hao Qinfen.Fast Benes network that supports fast low-cost partial route reconfiguration[J].Telecommunications Science,2026,42(01):86-104. DOI： 10.11959/j.issn.1000-0801.2026003.

摘要

为了解决互连规模大于100时快速可重构光互连网络单次局部重构代价过高的问题，提出了快速Benes网络与配套局部重构算法，利用预留空置链路减少局部重构对已有链路的影响，在互连规模超过100时性能优异。在处理单一节点的路由变更时，快速Benes网络仅影响平均2~4个接入节点对应的既有通信链路，略差于Crossbar网络，而远好于Benes网络（一次平均影响0.71

个接入节点，

为互连规模），降低重构代价达98%。基于该算法的现场可编程门阵列（field-programmable gate array，FPGA）硬件加速器，局部路由求解速度为79 ns/次，与Crossbar网络相近，比Benes网络快2个数量级。

Abstract

To address the problem of excessively high costs for single partial reconfiguration in a fast reconfigurable optical interconnection network when the interconnection scale is greater than 100

a fast Benes network and a supporting partial reconfiguration algorithm were proposed. By utilizing reserved idle links

the impact of partial reconfiguration on existing links was reduced

and it exhibited excellent performance when the interconnection scale exceeded 100. When dealing with the routing change of a single node

the existing communication links corresponding were affected to an average of 2~4 access nodes by the fast Benes network. It was only slightly inferior to the Crossbar network

and was much better than the Benes network. The reconfiguration cost was significantly reduced by up to 98%. Based on the FPGA hardware accelerator of this algorithm

the partial routing solution speed was 79 nanoseconds per time

which was similar to that of the Crossbar network and two orders of magnitude faster than the Benes network.

关键词

Keywords

references

Parsonson C W F , Shabka Z , Chlupka W K , et al . Optimal control of SOAs with artificial intelligence for sub-nanosecond optical switching [J ] . Journal of Lightwave Technology , 2020 , 38 ( 20 ): 5563 - 5573 .

Maniotis P , Dupuis N , Schares L , et al . Intra-node high-performance computing network architecture with nanosecond-scale photonic switches [J ] . Journal of Optical Communications and Networking , 2020 , 12 ( 12 ): 367 .

Farrington N , Porter G , Radhakrishnan S , et al . Helios: a hybrid electrical/optical switch architecture for modular data centers [C ] // Proceedings of the ACM SIGCOMM 2010 Conference . New York : ACM Press , 2010 : 339 - 350 .

Wang G H , Andersen D G , Kaminsky M , et al . C-Through: part-time optics in data centers [C ] // Proceedings of the ACM SIGCOMM 2010 Conference . New York : ACM Press , 2010 : 327 - 338 .

Bazzaz H H , Tewari M , Wang G H , et al . Switching the optical divide: fundamental challenges for hybrid electrical/optical datacenter networks [C ] // Proceedings of the 2nd ACM Symposium on Cloud Computing . New York : ACM Press , 2011 : 1 - 8 .

Beneš V E . Permutation groups, complexes, and rearrangeable connecting networks [J ] . Bell System Technical Journal , 1964 , 43 ( 4 ): 1619 - 1640 .

Zhao S Y , Lu L J , Zhou L J , et al . 16×16 silicon Mach–Zehnder interferometer switch actuated with waveguide microheaters [J ] . Photonics Research , 2016 , 4 ( 5 ): 202 .

Lu L J , Zhao S Y , Zhou L J , et al . 16×16 non-blocking silicon optical switch based on electro-optic Mach-Zehnder interferometers [J ] . Optics Express , 2016 , 24 ( 9 ): 9295 - 9307 .

Qiao L , Tang W J , Chu T . 32×32 silicon electro-optic switch with built-in monitors and balanced-status units [J ] . Scientific Reports , 2017 ( 7 ): 42306 .

Hou W G , Guo P X , Guo L , et al . O-star: an optical switching architecture featuring mode and wavelength-division multiplexing for on-chip many-core systems [J ] . Journal of Lightwave Technology , 2022 , 40 ( 1 ): 24 - 36 .

Wang Y , Qin Y B , Deng D Z , et al . A 28 nm 27.5 TOPS/W approximate-computing-based transformer processor with asymptotic sparsity speculating and out-of-order computing [C ] // Proceedings of the 2022 IEEE International Solid-State Circuits Conference (ISSCC) . Piscataway : IEEE Press , 2022 : 1 - 3 .

Waksman A . A permutation network [J ] . Journal of the ACM , 1968 , 15 ( 1 ): 159 - 163 .

Nassimi , Sahni . Parallel algorithms to set up the Benes permutation network [J ] . IEEE Transactions on Computers , 1982 , C-31( 2 ): 148 - 154 .

Lee K Y . A new benes network control algorithm [J ] . IEEE Transactions on Computers , 1987 , C-36( 6 ): 768 - 772 .

Koloko L , Matsumoto T , Obara H . Design and implementation of fast and hardware-efficient parallel processing elements to set full and partial permutations in Beneš networks [J ] . The Journal of Engineering , 2021 ( 6 ): 312 - 320 .

Nikolaidis D , Groumas P , Kouloumentas C , et al . Novel Benes network routing algorithm and hardware implementation [J ] . Technologies , 2022 , 10 ( 1 ): 16 .

秦梦远 , 刘宏伟 , 郝沁汾 . 高性能Benes网络路由求解算法及硬件加速器 [J ] . 计算机工程与应用 , 2025 , 61 ( 14 ): 163 - 175 .

Qin M Y , Liu H W , Hao Q F . High-performance Benes network routing algorithm and hardware accelerator [J ] . Computer Engineering and Applications , 2025 , 61 ( 14 ): 163 - 175 .

Wang C , Yoshikane N , Elson D , et al . Modoru: Clos nanosecond optical switching for distributed deep training [J ] . Journal of Optical Communications and Networking , 2024 , 16 ( 1 ): A40 - A52 .

Padmanabhan K , Netravali A . Dilated networks for photonic switching [J ] . IEEE Transactions on Communications , 1987 , 35 ( 12 ): 1357 - 1365 .

Kabacinski W . Modified dilated Benes networks for photonic switching [J ] . IEEE Transactions on Communications , 1999 , 47 ( 8 ): 1253 - 1259 .

Lu E Y , Zheng S Q . Fast reconfiguration algorithms for time, space, and wavelength dilated optical Benes networks [J ] . International Journal of Parallel, Emergent and Distributed Systems , 2007 , 22 ( 1 ): 39 - 58 .

张金花 , 武保剑 , 邱昆 . 扩张型Benes光交换集成芯片路由算法 [J ] . 光通信技术 , 2019 , 43 ( 2 ): 1 - 8 .

Zhang J H , Wu B J , Qiu K . Routing algorithm of dilated Benes optical switching integrated chip [J ] . Optical Communication Technology , 2019 , 43 ( 2 ): 1 - 8 .

张金花 , 武保剑 , 邱昆 . 扩张型Benes光交换芯片未满配置情形下的约束链路路由算法 [J ] . 激光与光电子学进展 , 2019 , 56 ( 21 ): 211301 .

Zhang J H , Wu B J , Qiu K . Constrained link routing algorithm for dilated Benes optical switching chips under non-full configuration [J ] . Laser & Optoelectronics Progress , 2019 , 56 ( 21 ): 211301 .

Goke L R , Lipovski G J . Banyan networks for partitioning multiprocessor systems [C ] // Proceedings of the 1st Annual Symposium on Computer Architecture - ISCA '73 . New York : ACM Press , 1973 : 21 - 28 .

Clos C . A study of non-blocking switching networks [J ] . Bell System Technical Journal , 1953 , 32 ( 2 ): 406 - 424 .

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

暂无数据