Network coding in networks-on-chip with lossy links

Michael Vonbun, Nguyen Anh Vu Doan, Thomas Wild, Andreas Herkersdorf

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

2 Scopus citations

Abstract

Providing a reliable and efficient communication infrastructure becomes more important and more demanding for current and future Many-core System-on-Chips. The increased amount of communication partners demands for efficient resource usage. At the same time, an on-going feature size reduction and supply-voltage level reduction renders the SoC vulnerable to bit errors. Our work aims at providing a generalized analysis of network-coded Networks-on-Chip with multicast support and unreliable links. We compare network-coded with dimension-routed NoCs in terms of their hop counts and achievable long-term rates. The results show that for an 8× 8 2D-mesh setup, network-coding yields a 1.7× to 8.3× better hop count and can achieve a 1.2× to 3.7× higher rate as dimension-routed 8 × 8 NoCs for bit error probabilities of 10-12 to 10-3 and 10% multicast traffic.

Original languageEnglish
Title of host publicationArchitecture of Computing Systems - ARCS 2019 - 32nd International Conference, Proceedings
EditorsChristian Hochberger, Martin Schoeberl, Jürgen Brehm, Thilo Pionteck, Sascha Uhrig
PublisherSpringer Verlag
Pages308-321
Number of pages14
ISBN (Print)9783030186555
DOIs
StatePublished - 2019
Event32nd International Conference on Architecture of Computing Systems, ARCS 2019 - Copenhagen, Denmark
Duration: 20 May 201923 May 2019

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11479 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference32nd International Conference on Architecture of Computing Systems, ARCS 2019
Country/TerritoryDenmark
CityCopenhagen
Period20/05/1923/05/19

Fingerprint

Dive into the research topics of 'Network coding in networks-on-chip with lossy links'. Together they form a unique fingerprint.

Cite this