Improving the performance of bristled CC-NUMA systems using virtual channels and adaptivity

Jose F. Martinez, Josep Torrellas, Jos Duato

Research output: Contribution to conferencePaper

Abstract

Current high-end parallel systems achieve low-latency, high-bandwidth network communication through the use of aggressive design techniques and expensive mechanical and electrical parts. High-speed interconnection networks, which are crucial to achieve acceptable system performance, may account for an important fraction of the total cost of the machine. To reduce the network cost and still maintain scalability, bristled configurations, in which each router connects to several processing nodes, pose an attractive alternative. Their lower bandwidth, however, may adversely affect the efficiency of the parallel codes. In this paper, we show how virtual channels and adaptive routing can make bristled systems more attractive: overall performance improves in congested scenarios while remaining practically unaltered under light traffic conditions. Experimental results are obtained by using execution-driven simulation of a complete state-of-the-art CC-NUMA system, with dynamic superscalar processors and contemporary pipelined routers. The results show that, in bristled hypercubes with 2 processing nodes per router, SPLASH-2 applications with significant communication run 5-15% faster if we make use of virtual channels and adaptive routing. The resulting systems are only 1-10% slower than systems with non-bristled hypercubes and similar routing support, even though the former only need about half of the network hardware components present in the latter. Additionally, virtual channels and adaptivity are shown to be of negligible effect in non-bristled hypercubes.

Original languageEnglish (US)
Pages202-209
Number of pages8
StatePublished - Jan 1 1999
EventProceedings of the 1999 13th ACM International Conference on Supercomputing, ICS'99 - Rhodes, Greece
Duration: Jun 20 1999Jun 25 1999

Other

OtherProceedings of the 1999 13th ACM International Conference on Supercomputing, ICS'99
CityRhodes, Greece
Period6/20/996/25/99

Fingerprint

Routers
Bandwidth
Processing
Telecommunication networks
Scalability
Costs
Hardware
Communication

ASJC Scopus subject areas

  • Computer Science(all)

Cite this

Martinez, J. F., Torrellas, J., & Duato, J. (1999). Improving the performance of bristled CC-NUMA systems using virtual channels and adaptivity. 202-209. Paper presented at Proceedings of the 1999 13th ACM International Conference on Supercomputing, ICS'99, Rhodes, Greece, .

Improving the performance of bristled CC-NUMA systems using virtual channels and adaptivity. / Martinez, Jose F.; Torrellas, Josep; Duato, Jos.

1999. 202-209 Paper presented at Proceedings of the 1999 13th ACM International Conference on Supercomputing, ICS'99, Rhodes, Greece, .

Research output: Contribution to conferencePaper

Martinez, JF, Torrellas, J & Duato, J 1999, 'Improving the performance of bristled CC-NUMA systems using virtual channels and adaptivity' Paper presented at Proceedings of the 1999 13th ACM International Conference on Supercomputing, ICS'99, Rhodes, Greece, 6/20/99 - 6/25/99, pp. 202-209.
Martinez JF, Torrellas J, Duato J. Improving the performance of bristled CC-NUMA systems using virtual channels and adaptivity. 1999. Paper presented at Proceedings of the 1999 13th ACM International Conference on Supercomputing, ICS'99, Rhodes, Greece, .
Martinez, Jose F. ; Torrellas, Josep ; Duato, Jos. / Improving the performance of bristled CC-NUMA systems using virtual channels and adaptivity. Paper presented at Proceedings of the 1999 13th ACM International Conference on Supercomputing, ICS'99, Rhodes, Greece, .8 p.
@conference{d03a79f1e3004c0a9d4c3dfff464383e,
title = "Improving the performance of bristled CC-NUMA systems using virtual channels and adaptivity",
abstract = "Current high-end parallel systems achieve low-latency, high-bandwidth network communication through the use of aggressive design techniques and expensive mechanical and electrical parts. High-speed interconnection networks, which are crucial to achieve acceptable system performance, may account for an important fraction of the total cost of the machine. To reduce the network cost and still maintain scalability, bristled configurations, in which each router connects to several processing nodes, pose an attractive alternative. Their lower bandwidth, however, may adversely affect the efficiency of the parallel codes. In this paper, we show how virtual channels and adaptive routing can make bristled systems more attractive: overall performance improves in congested scenarios while remaining practically unaltered under light traffic conditions. Experimental results are obtained by using execution-driven simulation of a complete state-of-the-art CC-NUMA system, with dynamic superscalar processors and contemporary pipelined routers. The results show that, in bristled hypercubes with 2 processing nodes per router, SPLASH-2 applications with significant communication run 5-15{\%} faster if we make use of virtual channels and adaptive routing. The resulting systems are only 1-10{\%} slower than systems with non-bristled hypercubes and similar routing support, even though the former only need about half of the network hardware components present in the latter. Additionally, virtual channels and adaptivity are shown to be of negligible effect in non-bristled hypercubes.",
author = "Martinez, {Jose F.} and Josep Torrellas and Jos Duato",
year = "1999",
month = "1",
day = "1",
language = "English (US)",
pages = "202--209",
note = "Proceedings of the 1999 13th ACM International Conference on Supercomputing, ICS'99 ; Conference date: 20-06-1999 Through 25-06-1999",

}

TY - CONF

T1 - Improving the performance of bristled CC-NUMA systems using virtual channels and adaptivity

AU - Martinez, Jose F.

AU - Torrellas, Josep

AU - Duato, Jos

PY - 1999/1/1

Y1 - 1999/1/1

N2 - Current high-end parallel systems achieve low-latency, high-bandwidth network communication through the use of aggressive design techniques and expensive mechanical and electrical parts. High-speed interconnection networks, which are crucial to achieve acceptable system performance, may account for an important fraction of the total cost of the machine. To reduce the network cost and still maintain scalability, bristled configurations, in which each router connects to several processing nodes, pose an attractive alternative. Their lower bandwidth, however, may adversely affect the efficiency of the parallel codes. In this paper, we show how virtual channels and adaptive routing can make bristled systems more attractive: overall performance improves in congested scenarios while remaining practically unaltered under light traffic conditions. Experimental results are obtained by using execution-driven simulation of a complete state-of-the-art CC-NUMA system, with dynamic superscalar processors and contemporary pipelined routers. The results show that, in bristled hypercubes with 2 processing nodes per router, SPLASH-2 applications with significant communication run 5-15% faster if we make use of virtual channels and adaptive routing. The resulting systems are only 1-10% slower than systems with non-bristled hypercubes and similar routing support, even though the former only need about half of the network hardware components present in the latter. Additionally, virtual channels and adaptivity are shown to be of negligible effect in non-bristled hypercubes.

AB - Current high-end parallel systems achieve low-latency, high-bandwidth network communication through the use of aggressive design techniques and expensive mechanical and electrical parts. High-speed interconnection networks, which are crucial to achieve acceptable system performance, may account for an important fraction of the total cost of the machine. To reduce the network cost and still maintain scalability, bristled configurations, in which each router connects to several processing nodes, pose an attractive alternative. Their lower bandwidth, however, may adversely affect the efficiency of the parallel codes. In this paper, we show how virtual channels and adaptive routing can make bristled systems more attractive: overall performance improves in congested scenarios while remaining practically unaltered under light traffic conditions. Experimental results are obtained by using execution-driven simulation of a complete state-of-the-art CC-NUMA system, with dynamic superscalar processors and contemporary pipelined routers. The results show that, in bristled hypercubes with 2 processing nodes per router, SPLASH-2 applications with significant communication run 5-15% faster if we make use of virtual channels and adaptive routing. The resulting systems are only 1-10% slower than systems with non-bristled hypercubes and similar routing support, even though the former only need about half of the network hardware components present in the latter. Additionally, virtual channels and adaptivity are shown to be of negligible effect in non-bristled hypercubes.

UR - http://www.scopus.com/inward/record.url?scp=0032662280&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0032662280&partnerID=8YFLogxK

M3 - Paper

AN - SCOPUS:0032662280

SP - 202

EP - 209

ER -