Contention-free many-to-many communication scheduling for high performance clusters

Satyajit Banerjee, Atish Datta Chowdhury, Koushik Sinha, Subhas Kumar Ghosh

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

4 Citations (Scopus)

Abstract

In the context of generating efficient, contention free schedules for inter-node communication through a switch fabric in cluster computing or data center type environments, all-to-all scheduling with equal sized data transfer requests has been studied in the literature [1,3,4]. In this paper, we propose a communication scheduling module (CSM) towards generating contention free communication schedules for many-to-many communication with arbitrary sized data. Towards this end, we propose three approximation algorithms - PST, LDT and SDT. From time to time, the CSM first generates a bipartite graph from the set of received requests, then determines which of these three algorithms gives the best approximation factor on this graph and finally executes that algorithm to generate a contention free schedule. Algorithm PST has a worst case run time of O( max (Δ|E|, |E|log(|E|))) and guarantees an approximation factor of 2H 2Δ-1, where |E| is the number of edges in the bipartite graph, Δ is the maximum node degree of the bipartite graph and H 2Δ-1 is the (2Δ-1)-th harmonic number. LDT runs in O(|E|2) and has an approximation factor of 2(1+τ), where τ is a constant defined as a guard band or pause time to eliminate the possibility of contention (in an apparently contention free schedule) caused by system jitter and synchronization inaccuracies between the nodes. SDT gives an approximation factor of 4log(w max ) and has a worst case run time of O(Δ|E|log(w max )), where w max represents the longest communication time in a set of received requests.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages150-161
Number of pages12
Volume6536 LNCS
DOIs
Publication statusPublished - 21 Feb 2011
Externally publishedYes
Event7th International Conference on Distributed Computing and Internet Technology, ICDCIT 2011 - Bhubaneshwar, India
Duration: 9 Feb 201112 Feb 2011

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume6536 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other7th International Conference on Distributed Computing and Internet Technology, ICDCIT 2011
CountryIndia
CityBhubaneshwar
Period9/2/1112/2/11

Fingerprint

Many to many
Contention
High Performance
Scheduling
Communication
Schedule
Bipartite Graph
Approximation
Vertex of a graph
Harmonic number
Cluster computing
Cluster Computing
Module
Data Center
Jitter
Data Transfer
Approximation algorithms
Data transfer
Best Approximation
Approximation Algorithms

Keywords

  • approximation algorithms
  • cluster computing
  • contention free schedule
  • data centers
  • Many-to-many scheduling
  • switch fabric
  • switch scheduling

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Banerjee, S., Datta Chowdhury, A., Sinha, K., & Ghosh, S. K. (2011). Contention-free many-to-many communication scheduling for high performance clusters. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 6536 LNCS, pp. 150-161). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6536 LNCS). https://doi.org/10.1007/978-3-642-19056-8_10

Contention-free many-to-many communication scheduling for high performance clusters. / Banerjee, Satyajit; Datta Chowdhury, Atish; Sinha, Koushik; Ghosh, Subhas Kumar.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 6536 LNCS 2011. p. 150-161 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6536 LNCS).

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

Banerjee, S, Datta Chowdhury, A, Sinha, K & Ghosh, SK 2011, Contention-free many-to-many communication scheduling for high performance clusters. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 6536 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 6536 LNCS, pp. 150-161, 7th International Conference on Distributed Computing and Internet Technology, ICDCIT 2011, Bhubaneshwar, India, 9/2/11. https://doi.org/10.1007/978-3-642-19056-8_10
Banerjee S, Datta Chowdhury A, Sinha K, Ghosh SK. Contention-free many-to-many communication scheduling for high performance clusters. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 6536 LNCS. 2011. p. 150-161. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-642-19056-8_10
Banerjee, Satyajit ; Datta Chowdhury, Atish ; Sinha, Koushik ; Ghosh, Subhas Kumar. / Contention-free many-to-many communication scheduling for high performance clusters. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 6536 LNCS 2011. pp. 150-161 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{4aeda75ae3e14860a8e53796ffe30591,
title = "Contention-free many-to-many communication scheduling for high performance clusters",
abstract = "In the context of generating efficient, contention free schedules for inter-node communication through a switch fabric in cluster computing or data center type environments, all-to-all scheduling with equal sized data transfer requests has been studied in the literature [1,3,4]. In this paper, we propose a communication scheduling module (CSM) towards generating contention free communication schedules for many-to-many communication with arbitrary sized data. Towards this end, we propose three approximation algorithms - PST, LDT and SDT. From time to time, the CSM first generates a bipartite graph from the set of received requests, then determines which of these three algorithms gives the best approximation factor on this graph and finally executes that algorithm to generate a contention free schedule. Algorithm PST has a worst case run time of O( max (Δ|E|, |E|log(|E|))) and guarantees an approximation factor of 2H 2Δ-1, where |E| is the number of edges in the bipartite graph, Δ is the maximum node degree of the bipartite graph and H 2Δ-1 is the (2Δ-1)-th harmonic number. LDT runs in O(|E|2) and has an approximation factor of 2(1+τ), where τ is a constant defined as a guard band or pause time to eliminate the possibility of contention (in an apparently contention free schedule) caused by system jitter and synchronization inaccuracies between the nodes. SDT gives an approximation factor of 4log(w max ) and has a worst case run time of O(Δ|E|log(w max )), where w max represents the longest communication time in a set of received requests.",
keywords = "approximation algorithms, cluster computing, contention free schedule, data centers, Many-to-many scheduling, switch fabric, switch scheduling",
author = "Satyajit Banerjee and {Datta Chowdhury}, Atish and Koushik Sinha and Ghosh, {Subhas Kumar}",
year = "2011",
month = "2",
day = "21",
doi = "10.1007/978-3-642-19056-8_10",
language = "English",
isbn = "9783642190551",
volume = "6536 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "150--161",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - Contention-free many-to-many communication scheduling for high performance clusters

AU - Banerjee, Satyajit

AU - Datta Chowdhury, Atish

AU - Sinha, Koushik

AU - Ghosh, Subhas Kumar

PY - 2011/2/21

Y1 - 2011/2/21

N2 - In the context of generating efficient, contention free schedules for inter-node communication through a switch fabric in cluster computing or data center type environments, all-to-all scheduling with equal sized data transfer requests has been studied in the literature [1,3,4]. In this paper, we propose a communication scheduling module (CSM) towards generating contention free communication schedules for many-to-many communication with arbitrary sized data. Towards this end, we propose three approximation algorithms - PST, LDT and SDT. From time to time, the CSM first generates a bipartite graph from the set of received requests, then determines which of these three algorithms gives the best approximation factor on this graph and finally executes that algorithm to generate a contention free schedule. Algorithm PST has a worst case run time of O( max (Δ|E|, |E|log(|E|))) and guarantees an approximation factor of 2H 2Δ-1, where |E| is the number of edges in the bipartite graph, Δ is the maximum node degree of the bipartite graph and H 2Δ-1 is the (2Δ-1)-th harmonic number. LDT runs in O(|E|2) and has an approximation factor of 2(1+τ), where τ is a constant defined as a guard band or pause time to eliminate the possibility of contention (in an apparently contention free schedule) caused by system jitter and synchronization inaccuracies between the nodes. SDT gives an approximation factor of 4log(w max ) and has a worst case run time of O(Δ|E|log(w max )), where w max represents the longest communication time in a set of received requests.

AB - In the context of generating efficient, contention free schedules for inter-node communication through a switch fabric in cluster computing or data center type environments, all-to-all scheduling with equal sized data transfer requests has been studied in the literature [1,3,4]. In this paper, we propose a communication scheduling module (CSM) towards generating contention free communication schedules for many-to-many communication with arbitrary sized data. Towards this end, we propose three approximation algorithms - PST, LDT and SDT. From time to time, the CSM first generates a bipartite graph from the set of received requests, then determines which of these three algorithms gives the best approximation factor on this graph and finally executes that algorithm to generate a contention free schedule. Algorithm PST has a worst case run time of O( max (Δ|E|, |E|log(|E|))) and guarantees an approximation factor of 2H 2Δ-1, where |E| is the number of edges in the bipartite graph, Δ is the maximum node degree of the bipartite graph and H 2Δ-1 is the (2Δ-1)-th harmonic number. LDT runs in O(|E|2) and has an approximation factor of 2(1+τ), where τ is a constant defined as a guard band or pause time to eliminate the possibility of contention (in an apparently contention free schedule) caused by system jitter and synchronization inaccuracies between the nodes. SDT gives an approximation factor of 4log(w max ) and has a worst case run time of O(Δ|E|log(w max )), where w max represents the longest communication time in a set of received requests.

KW - approximation algorithms

KW - cluster computing

KW - contention free schedule

KW - data centers

KW - Many-to-many scheduling

KW - switch fabric

KW - switch scheduling

UR - http://www.scopus.com/inward/record.url?scp=79951638887&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=79951638887&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-19056-8_10

DO - 10.1007/978-3-642-19056-8_10

M3 - Conference contribution

SN - 9783642190551

VL - 6536 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 150

EP - 161

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

ER -