Abstract

In existing multicomputers, software overhead dom inates the message-passing latency cost. Our research on the SHRIMP project at Princeton indicates that appropriate network interface support can significantly reduce this software overhead. We have designed two network interfaces for the SHRIMP multicomputer. Both support virtual memory mapped communication allowing user processes to communicate without doing expensive buffer management, and without using sys tem calls to cross the protection boundary separating user processes from the operating system kernel. This paper describes and compares the two network in terfaces, and discusses performance tradeoffs between them.

Original languageEnglish (US)
Article number765350
Pages (from-to)209-217
Number of pages9
JournalProceedings - Symposium on the High Performance Interconnects, Hot Interconnects
DOIs
StatePublished - Jan 1 1994
Event2nd Hot Interconnects Symposium, CONECT 1994 - Stanford, United States
Duration: Aug 11 1994Aug 13 1994

Fingerprint

Interfaces (computer)
Data storage equipment
Message passing
Communication
Costs

All Science Journal Classification (ASJC) codes

  • Engineering(all)

Cite this

@article{c9c573b3f69e4914beb281e48da52f13,
title = "Two virtual memory mapped network interface designs",
abstract = "In existing multicomputers, software overhead dom inates the message-passing latency cost. Our research on the SHRIMP project at Princeton indicates that appropriate network interface support can significantly reduce this software overhead. We have designed two network interfaces for the SHRIMP multicomputer. Both support virtual memory mapped communication allowing user processes to communicate without doing expensive buffer management, and without using sys tem calls to cross the protection boundary separating user processes from the operating system kernel. This paper describes and compares the two network in terfaces, and discusses performance tradeoffs between them.",
author = "Blumrich, {M. A.} and C. Dubnickil and Felten, {Edward William} and Kai Li and Mesarina, {M. R.}",
year = "1994",
month = "1",
day = "1",
doi = "https://doi.org/10.1109/CONECT.1994.765350",
language = "English (US)",
pages = "209--217",
journal = "Proceedings - Symposium on the High Performance Interconnects, Hot Interconnects",
issn = "1550-4794",

}

Two virtual memory mapped network interface designs. / Blumrich, M. A.; Dubnickil, C.; Felten, Edward William; Li, Kai; Mesarina, M. R.

In: Proceedings - Symposium on the High Performance Interconnects, Hot Interconnects, 01.01.1994, p. 209-217.

Research output: Contribution to journalConference article

TY - JOUR

T1 - Two virtual memory mapped network interface designs

AU - Blumrich, M. A.

AU - Dubnickil, C.

AU - Felten, Edward William

AU - Li, Kai

AU - Mesarina, M. R.

PY - 1994/1/1

Y1 - 1994/1/1

N2 - In existing multicomputers, software overhead dom inates the message-passing latency cost. Our research on the SHRIMP project at Princeton indicates that appropriate network interface support can significantly reduce this software overhead. We have designed two network interfaces for the SHRIMP multicomputer. Both support virtual memory mapped communication allowing user processes to communicate without doing expensive buffer management, and without using sys tem calls to cross the protection boundary separating user processes from the operating system kernel. This paper describes and compares the two network in terfaces, and discusses performance tradeoffs between them.

AB - In existing multicomputers, software overhead dom inates the message-passing latency cost. Our research on the SHRIMP project at Princeton indicates that appropriate network interface support can significantly reduce this software overhead. We have designed two network interfaces for the SHRIMP multicomputer. Both support virtual memory mapped communication allowing user processes to communicate without doing expensive buffer management, and without using sys tem calls to cross the protection boundary separating user processes from the operating system kernel. This paper describes and compares the two network in terfaces, and discusses performance tradeoffs between them.

UR - http://www.scopus.com/inward/record.url?scp=4544356240&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=4544356240&partnerID=8YFLogxK

U2 - https://doi.org/10.1109/CONECT.1994.765350

DO - https://doi.org/10.1109/CONECT.1994.765350

M3 - Conference article

SP - 209

EP - 217

JO - Proceedings - Symposium on the High Performance Interconnects, Hot Interconnects

JF - Proceedings - Symposium on the High Performance Interconnects, Hot Interconnects

SN - 1550-4794

M1 - 765350

ER -