Performance enhancement of SMP clusters with multiple network interfaces using virtualization

Peter Strazdins; Richard Alexander; David Barr

doi:10.1007/11942634_47

Performance enhancement of SMP clusters with multiple network interfaces using virtualization

Peter Strazdins, Richard Alexander, David Barr

Research output: Chapter in Book/Report/Conference proceeding › Conference Paper › peer-review

5 Citations (Scopus)

Abstract

Clusters of small-scale SMP/CMP nodes are becoming increasingly popular due to their cost-effectiveness. As these nodes are typically capable of supporting a number of network interfaces similar to the number of CPUs, the issue arises how to optimally configure the cluster for optimum communication performance. This paper evaluates a number of configurations on a 4-CPU Opteron cluster with multiple Gigabit Ethernet interfaces. Techniques include channel bonding and using independent communication pathways. With the latter, the use of virtualization via the Xen Virtual Machine Monitor offers the best potential to parallelize all stages of message transmission, for the case when multiple CPUs on a node are communicating simultaneously. Network-level microbenchmarks indicate the best performance is achieved with a configuration where guest virtual machines running on each CPU communicate directly with a dedicated interface, bypassing the virtual machine monitor. Channel bonding also proved to be more effective over multiple communication streams than over single.

Original language	English
Title of host publication	Frontiers of High Performance Computing and Networking - ISPA 2006 International Workshops, FHPCN, XHPC, S-GRACE, GridGIS, HPC-GTP, PDCE, ParDMCom, WOMP, ISDF, and UPWN, Proceedings
Pages	452-463
Number of pages	12
DOIs	https://doi.org/10.1007/11942634_47
Publication status	Published - 2006
Event	Int. Workshops on FHPCN 2006, XHPC 2006, S-GRACE 2006, GridGIS 2006, HPC-GTP 2006, PDCE 2006, ParDMCom 2006, WOMP 2006, ISDF 2006, and UPWN 2006, Held in Conjunction with the 4th Int. Symp. on Parallel and Distributed Processing and Appl., SPA 2006 - Sorrento, Italy Duration: 4 Dec 2006 → 7 Dec 2006

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	4331 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	Int. Workshops on FHPCN 2006, XHPC 2006, S-GRACE 2006, GridGIS 2006, HPC-GTP 2006, PDCE 2006, ParDMCom 2006, WOMP 2006, ISDF 2006, and UPWN 2006, Held in Conjunction with the 4th Int. Symp. on Parallel and Distributed Processing and Appl., SPA 2006
Country/Territory	Italy
City	Sorrento
Period	4/12/06 → 7/12/06

Access to Document

10.1007/11942634_47

Cite this

Strazdins, P., Alexander, R., & Barr, D. (2006). Performance enhancement of SMP clusters with multiple network interfaces using virtualization. In Frontiers of High Performance Computing and Networking - ISPA 2006 International Workshops, FHPCN, XHPC, S-GRACE, GridGIS, HPC-GTP, PDCE, ParDMCom, WOMP, ISDF, and UPWN, Proceedings (pp. 452-463). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 4331 LNCS). https://doi.org/10.1007/11942634_47

Strazdins, Peter ; Alexander, Richard ; Barr, David. / Performance enhancement of SMP clusters with multiple network interfaces using virtualization. Frontiers of High Performance Computing and Networking - ISPA 2006 International Workshops, FHPCN, XHPC, S-GRACE, GridGIS, HPC-GTP, PDCE, ParDMCom, WOMP, ISDF, and UPWN, Proceedings. 2006. pp. 452-463 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{49cc4c6e350e4ceeaf0c0d0ea379e3a9,

title = "Performance enhancement of SMP clusters with multiple network interfaces using virtualization",

abstract = "Clusters of small-scale SMP/CMP nodes are becoming increasingly popular due to their cost-effectiveness. As these nodes are typically capable of supporting a number of network interfaces similar to the number of CPUs, the issue arises how to optimally configure the cluster for optimum communication performance. This paper evaluates a number of configurations on a 4-CPU Opteron cluster with multiple Gigabit Ethernet interfaces. Techniques include channel bonding and using independent communication pathways. With the latter, the use of virtualization via the Xen Virtual Machine Monitor offers the best potential to parallelize all stages of message transmission, for the case when multiple CPUs on a node are communicating simultaneously. Network-level microbenchmarks indicate the best performance is achieved with a configuration where guest virtual machines running on each CPU communicate directly with a dedicated interface, bypassing the virtual machine monitor. Channel bonding also proved to be more effective over multiple communication streams than over single.",

author = "Peter Strazdins and Richard Alexander and David Barr",

year = "2006",

doi = "10.1007/11942634\_47",

language = "English",

isbn = "3540498605",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

pages = "452--463",

booktitle = "Frontiers of High Performance Computing and Networking - ISPA 2006 International Workshops, FHPCN, XHPC, S-GRACE, GridGIS, HPC-GTP, PDCE, ParDMCom, WOMP, ISDF, and UPWN, Proceedings",

note = "Int. Workshops on FHPCN 2006, XHPC 2006, S-GRACE 2006, GridGIS 2006, HPC-GTP 2006, PDCE 2006, ParDMCom 2006, WOMP 2006, ISDF 2006, and UPWN 2006, Held in Conjunction with the 4th Int. Symp. on Parallel and Distributed Processing and Appl., SPA 2006 ; Conference date: 04-12-2006 Through 07-12-2006",

}

Strazdins, P, Alexander, R & Barr, D 2006, Performance enhancement of SMP clusters with multiple network interfaces using virtualization. in Frontiers of High Performance Computing and Networking - ISPA 2006 International Workshops, FHPCN, XHPC, S-GRACE, GridGIS, HPC-GTP, PDCE, ParDMCom, WOMP, ISDF, and UPWN, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 4331 LNCS, pp. 452-463, Int. Workshops on FHPCN 2006, XHPC 2006, S-GRACE 2006, GridGIS 2006, HPC-GTP 2006, PDCE 2006, ParDMCom 2006, WOMP 2006, ISDF 2006, and UPWN 2006, Held in Conjunction with the 4th Int. Symp. on Parallel and Distributed Processing and Appl., SPA 2006, Sorrento, Italy, 4/12/06. https://doi.org/10.1007/11942634_47

Performance enhancement of SMP clusters with multiple network interfaces using virtualization. / Strazdins, Peter; Alexander, Richard; Barr, David.
Frontiers of High Performance Computing and Networking - ISPA 2006 International Workshops, FHPCN, XHPC, S-GRACE, GridGIS, HPC-GTP, PDCE, ParDMCom, WOMP, ISDF, and UPWN, Proceedings. 2006. p. 452-463 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 4331 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference Paper › peer-review

TY - GEN

T1 - Performance enhancement of SMP clusters with multiple network interfaces using virtualization

AU - Strazdins, Peter

AU - Alexander, Richard

AU - Barr, David

PY - 2006

Y1 - 2006

N2 - Clusters of small-scale SMP/CMP nodes are becoming increasingly popular due to their cost-effectiveness. As these nodes are typically capable of supporting a number of network interfaces similar to the number of CPUs, the issue arises how to optimally configure the cluster for optimum communication performance. This paper evaluates a number of configurations on a 4-CPU Opteron cluster with multiple Gigabit Ethernet interfaces. Techniques include channel bonding and using independent communication pathways. With the latter, the use of virtualization via the Xen Virtual Machine Monitor offers the best potential to parallelize all stages of message transmission, for the case when multiple CPUs on a node are communicating simultaneously. Network-level microbenchmarks indicate the best performance is achieved with a configuration where guest virtual machines running on each CPU communicate directly with a dedicated interface, bypassing the virtual machine monitor. Channel bonding also proved to be more effective over multiple communication streams than over single.

AB - Clusters of small-scale SMP/CMP nodes are becoming increasingly popular due to their cost-effectiveness. As these nodes are typically capable of supporting a number of network interfaces similar to the number of CPUs, the issue arises how to optimally configure the cluster for optimum communication performance. This paper evaluates a number of configurations on a 4-CPU Opteron cluster with multiple Gigabit Ethernet interfaces. Techniques include channel bonding and using independent communication pathways. With the latter, the use of virtualization via the Xen Virtual Machine Monitor offers the best potential to parallelize all stages of message transmission, for the case when multiple CPUs on a node are communicating simultaneously. Network-level microbenchmarks indicate the best performance is achieved with a configuration where guest virtual machines running on each CPU communicate directly with a dedicated interface, bypassing the virtual machine monitor. Channel bonding also proved to be more effective over multiple communication streams than over single.

UR - https://www.scopus.com/pages/publications/41549158916

U2 - 10.1007/11942634_47

DO - 10.1007/11942634_47

M3 - Conference Paper

SN - 3540498605

SN - 9783540498605

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

SP - 452

EP - 463

BT - Frontiers of High Performance Computing and Networking - ISPA 2006 International Workshops, FHPCN, XHPC, S-GRACE, GridGIS, HPC-GTP, PDCE, ParDMCom, WOMP, ISDF, and UPWN, Proceedings

T2 - Int. Workshops on FHPCN 2006, XHPC 2006, S-GRACE 2006, GridGIS 2006, HPC-GTP 2006, PDCE 2006, ParDMCom 2006, WOMP 2006, ISDF 2006, and UPWN 2006, Held in Conjunction with the 4th Int. Symp. on Parallel and Distributed Processing and Appl., SPA 2006

Y2 - 4 December 2006 through 7 December 2006

ER -

Strazdins P, Alexander R, Barr D. Performance enhancement of SMP clusters with multiple network interfaces using virtualization. In Frontiers of High Performance Computing and Networking - ISPA 2006 International Workshops, FHPCN, XHPC, S-GRACE, GridGIS, HPC-GTP, PDCE, ParDMCom, WOMP, ISDF, and UPWN, Proceedings. 2006. p. 452-463. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/11942634_47

Performance enhancement of SMP clusters with multiple network interfaces using virtualization

Abstract

Publication series

Conference

Access to Document

Other files and links

Fingerprint

Cite this