Optimisation of extended generalised fat tree topologies

Adamantini Peratikou, Mo Adda

Research output: Chapter in Book/Report/Conference proceedingConference contribution

161 Downloads (Pure)

Abstract

Extended generalised fat tree (XGFT) are interconnection networks with bidirectional multistage properties, (BMIN) which can be extended, or scaled to accommodate different system sizes and requirements. However, these extended topologies do not address power consumption and traffic constraints. In this paper, we extract a sub-set of the generalised fat tree topologies that are power consumption and performance aware. We called this sub-set optimised OXGFT. The cost which is proportional to the relative power is the objective function that is minimised based on the traffic constraints to maintain a lower delay and a higher throughput. The simulation results show that the extracted OXGFT topologies perform well under various load conditions.
Original languageEnglish
Title of host publicationDistributed computer and communication networks
Subtitle of host publication17th international conference, DCCN 2013, Moscow, Russia, October 7-10, 2013. revised selected papers
EditorsVladimir Vishnevsky, Dmitry Kozyrev, Andrey Larionov
Place of PublicationHeidelberg
PublisherSpringer
Pages82-90
ISBN (Electronic)9783319052090
ISBN (Print)9783319052083
DOIs
Publication statusPublished - 2014
Event17th International Conference on Distributed Computer and Communication Networks - Moscow, Russian Federation
Duration: 7 Oct 201310 Oct 2013

Publication series

NameCommunications in computer and information science
PublisherSpringer
Volume279
ISSN (Print)1865-0929

Conference

Conference17th International Conference on Distributed Computer and Communication Networks
Country/TerritoryRussian Federation
CityMoscow
Period7/10/1310/10/13

Fingerprint

Dive into the research topics of 'Optimisation of extended generalised fat tree topologies'. Together they form a unique fingerprint.

Cite this