Fine-grained system-call scheduling in CEFOS on commodity processors

Shigeru Kusakabe, Kentaro Iio, Hideo Taniguchi, Makoto Amamiya

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

We are developing an operating system, CEFOS, based on a fine-grain multi-threading model. By consisting a process with fine-grain threads, we can employ flexible scheduling schemes as long as thread dependencies allow. This paper presents WSC mechanism in CEFOS, which buffers multiple system-calls until the number of system-call requests or the time-slice reaches some threshold and then transfers the control to the operating system kernel with a bucket of the buffered system-call requests. We have developed CEFOS based on Linux and measured the system-call performance on commodity Intel processors. While fine-grained system-call scheduling with WSC caused frequent cache misses, it reduced context switches and improved the total execution time thanks to the reduction of TLB misses.

Original languageEnglish
Pages (from-to)383-388
Number of pages6
JournalLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume3019
Publication statusPublished - 2004
Externally publishedYes

Fingerprint

Scheduling
Operating Systems
Thread
Switches
Multithreading
Linux
Slice
Cache
Execution Time
Buffer
Switch
kernel
Model

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

@article{80e883b75f104df8a5ecbad5d856620b,
title = "Fine-grained system-call scheduling in CEFOS on commodity processors",
abstract = "We are developing an operating system, CEFOS, based on a fine-grain multi-threading model. By consisting a process with fine-grain threads, we can employ flexible scheduling schemes as long as thread dependencies allow. This paper presents WSC mechanism in CEFOS, which buffers multiple system-calls until the number of system-call requests or the time-slice reaches some threshold and then transfers the control to the operating system kernel with a bucket of the buffered system-call requests. We have developed CEFOS based on Linux and measured the system-call performance on commodity Intel processors. While fine-grained system-call scheduling with WSC caused frequent cache misses, it reduced context switches and improved the total execution time thanks to the reduction of TLB misses.",
author = "Shigeru Kusakabe and Kentaro Iio and Hideo Taniguchi and Makoto Amamiya",
year = "2004",
language = "English",
volume = "3019",
pages = "383--388",
journal = "Lecture Notes in Computer Science",
issn = "0302-9743",
publisher = "Springer Verlag",

}

TY - JOUR

T1 - Fine-grained system-call scheduling in CEFOS on commodity processors

AU - Kusakabe, Shigeru

AU - Iio, Kentaro

AU - Taniguchi, Hideo

AU - Amamiya, Makoto

PY - 2004

Y1 - 2004

N2 - We are developing an operating system, CEFOS, based on a fine-grain multi-threading model. By consisting a process with fine-grain threads, we can employ flexible scheduling schemes as long as thread dependencies allow. This paper presents WSC mechanism in CEFOS, which buffers multiple system-calls until the number of system-call requests or the time-slice reaches some threshold and then transfers the control to the operating system kernel with a bucket of the buffered system-call requests. We have developed CEFOS based on Linux and measured the system-call performance on commodity Intel processors. While fine-grained system-call scheduling with WSC caused frequent cache misses, it reduced context switches and improved the total execution time thanks to the reduction of TLB misses.

AB - We are developing an operating system, CEFOS, based on a fine-grain multi-threading model. By consisting a process with fine-grain threads, we can employ flexible scheduling schemes as long as thread dependencies allow. This paper presents WSC mechanism in CEFOS, which buffers multiple system-calls until the number of system-call requests or the time-slice reaches some threshold and then transfers the control to the operating system kernel with a bucket of the buffered system-call requests. We have developed CEFOS based on Linux and measured the system-call performance on commodity Intel processors. While fine-grained system-call scheduling with WSC caused frequent cache misses, it reduced context switches and improved the total execution time thanks to the reduction of TLB misses.

UR - http://www.scopus.com/inward/record.url?scp=33744777749&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33744777749&partnerID=8YFLogxK

M3 - Article

VL - 3019

SP - 383

EP - 388

JO - Lecture Notes in Computer Science

JF - Lecture Notes in Computer Science

SN - 0302-9743

ER -