Router-level performance driven dynamic management in hierarchical networks-on-chip

Mingmin Bai, Dan Zhao, M. Bayoumi
{"title":"Router-level performance driven dynamic management in hierarchical networks-on-chip","authors":"Mingmin Bai, Dan Zhao, M. Bayoumi","doi":"10.1109/SOCC.2017.8226067","DOIUrl":null,"url":null,"abstract":"When on-chip interconnection network scales to integrate more processing elements, the average end-to-end latency is highly increased due to long average hop distance. Though it has been discovered that, almost of the communication in large scale networks is between nodes in a short range, it revealed that the small portion of data delivery between distant nodes consumes or occupies most of the network bandwidth. Hierarchical NoCs caters an attractive solution to resolve the distant data transmission problem by taking advantage of the network hierarchy. However, it brings about new sever congestion challenge because of uneven traffic distribution among hierarchy. In previous work, we performed a detouring scheme on a layered hierarchical NoC. When congestion is formed on the access link to adjacent hierarchical layer, the detouring scheme seeks and reroutes the packets to an nearby node to access the next adjacent network layer. It revealed that the links, which bridges the packets up to higher layers, are more essential for distributing the traffic and avoiding congestion between hierarchy levels. In this paper, we proposed dynamic schemes to solve the congestion problem introduced by region-based hierarchical routing on a hierarchical NoC. The results exposed that the dynamic approaches are efficient to manage the congestion under heavier long range traffic load, yielding significant average network latency reduction and throughput increment under mixed synthetic traffic patterns.","PeriodicalId":366264,"journal":{"name":"2017 30th IEEE International System-on-Chip Conference (SOCC)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"3","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2017 30th IEEE International System-on-Chip Conference (SOCC)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/SOCC.2017.8226067","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 3

Abstract

When on-chip interconnection network scales to integrate more processing elements, the average end-to-end latency is highly increased due to long average hop distance. Though it has been discovered that, almost of the communication in large scale networks is between nodes in a short range, it revealed that the small portion of data delivery between distant nodes consumes or occupies most of the network bandwidth. Hierarchical NoCs caters an attractive solution to resolve the distant data transmission problem by taking advantage of the network hierarchy. However, it brings about new sever congestion challenge because of uneven traffic distribution among hierarchy. In previous work, we performed a detouring scheme on a layered hierarchical NoC. When congestion is formed on the access link to adjacent hierarchical layer, the detouring scheme seeks and reroutes the packets to an nearby node to access the next adjacent network layer. It revealed that the links, which bridges the packets up to higher layers, are more essential for distributing the traffic and avoiding congestion between hierarchy levels. In this paper, we proposed dynamic schemes to solve the congestion problem introduced by region-based hierarchical routing on a hierarchical NoC. The results exposed that the dynamic approaches are efficient to manage the congestion under heavier long range traffic load, yielding significant average network latency reduction and throughput increment under mixed synthetic traffic patterns.
分层片上网络中路由器级性能驱动的动态管理
当片上互连网络扩展到集成更多处理元素时,由于平均跳距较长,端到端平均延迟大大增加。虽然已经发现,在大规模网络中,几乎所有的通信都是在近距离的节点之间进行的,但这表明,远距离节点之间的一小部分数据传输消耗或占用了大部分网络带宽。分层noc利用网络分层的优势,为解决远程数据传输问题提供了一个有吸引力的解决方案。然而,由于层级之间的流量分布不均匀,给交通拥挤带来了新的严峻挑战。在之前的工作中,我们在分层分层NoC上执行了绕行方案。当到相邻分层层的访问链路上形成拥塞时,绕道方案寻找并重新路由数据包到附近的节点以访问下一个相邻的网络层。它揭示了连接数据包到更高层的链路对于分配流量和避免分层层之间的拥塞更为重要。本文提出了一种动态方案来解决分层NoC上基于区域的分层路由所带来的拥塞问题。结果表明,在较重的远程流量负载下,动态方法可以有效地管理拥塞,在混合综合流量模式下,平均网络延迟显著降低,吞吐量显著增加。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信