An application-driven study of parallel system overheads and network bandwidth requirements

Anand Sivasubramaniam, Aman Singla, Umakishore Ramachandran, H. Venkateswaran

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

Evaluating and analyzing the performance of a parallel application on an architecture to explain the disparity between projected and delivered performance is an important aspect of parallel systems research. However, conducting such a study is hard due to the vast design space of these systems. In this paper, we study two important aspects related to the performance of parallel applications on shared memory parallel architectures. First, we quantify overheads observed during the execution of these applications on three different simulated architectures. We next use these results to synthesize the bandwidth requirements for the applications with respect to different network topologies. This study is performed using an execution-driven simulation tool called SPASM, which provides a way of isolating and quantifying the different parallel system overheads in a nonintrusive manner. The first exercise shows that in shared memory machines with private caches, as long as the applications are well-structured to exploit locality, the key determinant that impacts performance is network connection. The second exercise quantifies the network bandwidth needed to minimize the effect of network connection. Specifically, it is shown that for the applications considered, as long as the problem sizes are increased commensurate with the system size, current network technologies supporting 200-300 MBytes/sec link bandwidth are sufficient to keep the network overheads (such as latency and contention) within acceptable bounds.

Original languageEnglish (US)
Pages (from-to)193-210
Number of pages18
JournalIEEE Transactions on Parallel and Distributed Systems
Volume10
Issue number3
DOIs
StatePublished - Dec 1 1999

Fingerprint

Bandwidth
Data storage equipment
Parallel architectures
Network performance
Telecommunication links
Topology

All Science Journal Classification (ASJC) codes

  • Signal Processing
  • Hardware and Architecture
  • Computational Theory and Mathematics

Cite this

@article{8a7f835906244996a89cb7f205600054,
title = "An application-driven study of parallel system overheads and network bandwidth requirements",
abstract = "Evaluating and analyzing the performance of a parallel application on an architecture to explain the disparity between projected and delivered performance is an important aspect of parallel systems research. However, conducting such a study is hard due to the vast design space of these systems. In this paper, we study two important aspects related to the performance of parallel applications on shared memory parallel architectures. First, we quantify overheads observed during the execution of these applications on three different simulated architectures. We next use these results to synthesize the bandwidth requirements for the applications with respect to different network topologies. This study is performed using an execution-driven simulation tool called SPASM, which provides a way of isolating and quantifying the different parallel system overheads in a nonintrusive manner. The first exercise shows that in shared memory machines with private caches, as long as the applications are well-structured to exploit locality, the key determinant that impacts performance is network connection. The second exercise quantifies the network bandwidth needed to minimize the effect of network connection. Specifically, it is shown that for the applications considered, as long as the problem sizes are increased commensurate with the system size, current network technologies supporting 200-300 MBytes/sec link bandwidth are sufficient to keep the network overheads (such as latency and contention) within acceptable bounds.",
author = "Anand Sivasubramaniam and Aman Singla and Umakishore Ramachandran and H. Venkateswaran",
year = "1999",
month = "12",
day = "1",
doi = "10.1109/71.755819",
language = "English (US)",
volume = "10",
pages = "193--210",
journal = "IEEE Transactions on Parallel and Distributed Systems",
issn = "1045-9219",
publisher = "IEEE Computer Society",
number = "3",

}

An application-driven study of parallel system overheads and network bandwidth requirements. / Sivasubramaniam, Anand; Singla, Aman; Ramachandran, Umakishore; Venkateswaran, H.

In: IEEE Transactions on Parallel and Distributed Systems, Vol. 10, No. 3, 01.12.1999, p. 193-210.

Research output: Contribution to journalArticle

TY - JOUR

T1 - An application-driven study of parallel system overheads and network bandwidth requirements

AU - Sivasubramaniam, Anand

AU - Singla, Aman

AU - Ramachandran, Umakishore

AU - Venkateswaran, H.

PY - 1999/12/1

Y1 - 1999/12/1

N2 - Evaluating and analyzing the performance of a parallel application on an architecture to explain the disparity between projected and delivered performance is an important aspect of parallel systems research. However, conducting such a study is hard due to the vast design space of these systems. In this paper, we study two important aspects related to the performance of parallel applications on shared memory parallel architectures. First, we quantify overheads observed during the execution of these applications on three different simulated architectures. We next use these results to synthesize the bandwidth requirements for the applications with respect to different network topologies. This study is performed using an execution-driven simulation tool called SPASM, which provides a way of isolating and quantifying the different parallel system overheads in a nonintrusive manner. The first exercise shows that in shared memory machines with private caches, as long as the applications are well-structured to exploit locality, the key determinant that impacts performance is network connection. The second exercise quantifies the network bandwidth needed to minimize the effect of network connection. Specifically, it is shown that for the applications considered, as long as the problem sizes are increased commensurate with the system size, current network technologies supporting 200-300 MBytes/sec link bandwidth are sufficient to keep the network overheads (such as latency and contention) within acceptable bounds.

AB - Evaluating and analyzing the performance of a parallel application on an architecture to explain the disparity between projected and delivered performance is an important aspect of parallel systems research. However, conducting such a study is hard due to the vast design space of these systems. In this paper, we study two important aspects related to the performance of parallel applications on shared memory parallel architectures. First, we quantify overheads observed during the execution of these applications on three different simulated architectures. We next use these results to synthesize the bandwidth requirements for the applications with respect to different network topologies. This study is performed using an execution-driven simulation tool called SPASM, which provides a way of isolating and quantifying the different parallel system overheads in a nonintrusive manner. The first exercise shows that in shared memory machines with private caches, as long as the applications are well-structured to exploit locality, the key determinant that impacts performance is network connection. The second exercise quantifies the network bandwidth needed to minimize the effect of network connection. Specifically, it is shown that for the applications considered, as long as the problem sizes are increased commensurate with the system size, current network technologies supporting 200-300 MBytes/sec link bandwidth are sufficient to keep the network overheads (such as latency and contention) within acceptable bounds.

UR - http://www.scopus.com/inward/record.url?scp=0032656539&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0032656539&partnerID=8YFLogxK

U2 - 10.1109/71.755819

DO - 10.1109/71.755819

M3 - Article

AN - SCOPUS:0032656539

VL - 10

SP - 193

EP - 210

JO - IEEE Transactions on Parallel and Distributed Systems

JF - IEEE Transactions on Parallel and Distributed Systems

SN - 1045-9219

IS - 3

ER -