Measurements of a distributed file system

Mary G. Baker, John H Hartman, Michael D. Kupfer, Ken W. Shirriff, John K. Ousterhout

Research output: Chapter in Book/Report/Conference proceedingConference contribution

269 Citations (Scopus)

Abstract

We analyzed the user-level file access patterns and caching behavior of the Sprite distributed file system. The first part of our analysis repeated a study done in 1985 of the: BSD UNIX file system. We found that file throughput has increased by a factor of 20 to an average of 8 Kbytes per second per active user over 10-minute intervals, and that the use of process migration for load sharing increased burst rates by another factor of six. Also, many more very large (multi-megabyte) files are in use today than in 1985. The second part of our analysis measured the behavior of Sprite's main-memory file caches. Client-level caches average about 7 Mbytes in size (about one-quarter to one-third of main memory) and filter out about 50% of the traffic between clients and servers. 35% of the remaining server traffic is caused by paging, even on workstations with large memories. We found that client cache consistency is needed to prevent stale data errors, but that it is not invoked often enough to degrade overall system performance.

Original languageEnglish (US)
Title of host publicationProceedings of the 13th ACM Symposium on Operating Systems Principles, SOSP 1991
Pages198-212
Number of pages15
DOIs
StatePublished - 1991
Externally publishedYes
Event13th ACM Symposium on Operating Systems Principles, SOSP 1991 - Pacific Grove, CA, United States
Duration: Oct 13 1991Oct 16 1991

Other

Other13th ACM Symposium on Operating Systems Principles, SOSP 1991
CountryUnited States
CityPacific Grove, CA
Period10/13/9110/16/91

Fingerprint

Data storage equipment
Servers
UNIX
Computer workstations
Throughput

ASJC Scopus subject areas

  • Software

Cite this

Baker, M. G., Hartman, J. H., Kupfer, M. D., Shirriff, K. W., & Ousterhout, J. K. (1991). Measurements of a distributed file system. In Proceedings of the 13th ACM Symposium on Operating Systems Principles, SOSP 1991 (pp. 198-212) https://doi.org/10.1145/121132.121164

Measurements of a distributed file system. / Baker, Mary G.; Hartman, John H; Kupfer, Michael D.; Shirriff, Ken W.; Ousterhout, John K.

Proceedings of the 13th ACM Symposium on Operating Systems Principles, SOSP 1991. 1991. p. 198-212.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Baker, MG, Hartman, JH, Kupfer, MD, Shirriff, KW & Ousterhout, JK 1991, Measurements of a distributed file system. in Proceedings of the 13th ACM Symposium on Operating Systems Principles, SOSP 1991. pp. 198-212, 13th ACM Symposium on Operating Systems Principles, SOSP 1991, Pacific Grove, CA, United States, 10/13/91. https://doi.org/10.1145/121132.121164
Baker MG, Hartman JH, Kupfer MD, Shirriff KW, Ousterhout JK. Measurements of a distributed file system. In Proceedings of the 13th ACM Symposium on Operating Systems Principles, SOSP 1991. 1991. p. 198-212 https://doi.org/10.1145/121132.121164
Baker, Mary G. ; Hartman, John H ; Kupfer, Michael D. ; Shirriff, Ken W. ; Ousterhout, John K. / Measurements of a distributed file system. Proceedings of the 13th ACM Symposium on Operating Systems Principles, SOSP 1991. 1991. pp. 198-212
@inproceedings{b75cc678625145cba1417962a376081e,
title = "Measurements of a distributed file system",
abstract = "We analyzed the user-level file access patterns and caching behavior of the Sprite distributed file system. The first part of our analysis repeated a study done in 1985 of the: BSD UNIX file system. We found that file throughput has increased by a factor of 20 to an average of 8 Kbytes per second per active user over 10-minute intervals, and that the use of process migration for load sharing increased burst rates by another factor of six. Also, many more very large (multi-megabyte) files are in use today than in 1985. The second part of our analysis measured the behavior of Sprite's main-memory file caches. Client-level caches average about 7 Mbytes in size (about one-quarter to one-third of main memory) and filter out about 50{\%} of the traffic between clients and servers. 35{\%} of the remaining server traffic is caused by paging, even on workstations with large memories. We found that client cache consistency is needed to prevent stale data errors, but that it is not invoked often enough to degrade overall system performance.",
author = "Baker, {Mary G.} and Hartman, {John H} and Kupfer, {Michael D.} and Shirriff, {Ken W.} and Ousterhout, {John K.}",
year = "1991",
doi = "10.1145/121132.121164",
language = "English (US)",
isbn = "0897914473",
pages = "198--212",
booktitle = "Proceedings of the 13th ACM Symposium on Operating Systems Principles, SOSP 1991",

}

TY - GEN

T1 - Measurements of a distributed file system

AU - Baker, Mary G.

AU - Hartman, John H

AU - Kupfer, Michael D.

AU - Shirriff, Ken W.

AU - Ousterhout, John K.

PY - 1991

Y1 - 1991

N2 - We analyzed the user-level file access patterns and caching behavior of the Sprite distributed file system. The first part of our analysis repeated a study done in 1985 of the: BSD UNIX file system. We found that file throughput has increased by a factor of 20 to an average of 8 Kbytes per second per active user over 10-minute intervals, and that the use of process migration for load sharing increased burst rates by another factor of six. Also, many more very large (multi-megabyte) files are in use today than in 1985. The second part of our analysis measured the behavior of Sprite's main-memory file caches. Client-level caches average about 7 Mbytes in size (about one-quarter to one-third of main memory) and filter out about 50% of the traffic between clients and servers. 35% of the remaining server traffic is caused by paging, even on workstations with large memories. We found that client cache consistency is needed to prevent stale data errors, but that it is not invoked often enough to degrade overall system performance.

AB - We analyzed the user-level file access patterns and caching behavior of the Sprite distributed file system. The first part of our analysis repeated a study done in 1985 of the: BSD UNIX file system. We found that file throughput has increased by a factor of 20 to an average of 8 Kbytes per second per active user over 10-minute intervals, and that the use of process migration for load sharing increased burst rates by another factor of six. Also, many more very large (multi-megabyte) files are in use today than in 1985. The second part of our analysis measured the behavior of Sprite's main-memory file caches. Client-level caches average about 7 Mbytes in size (about one-quarter to one-third of main memory) and filter out about 50% of the traffic between clients and servers. 35% of the remaining server traffic is caused by paging, even on workstations with large memories. We found that client cache consistency is needed to prevent stale data errors, but that it is not invoked often enough to degrade overall system performance.

UR - http://www.scopus.com/inward/record.url?scp=84883303431&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84883303431&partnerID=8YFLogxK

U2 - 10.1145/121132.121164

DO - 10.1145/121132.121164

M3 - Conference contribution

AN - SCOPUS:84883303431

SN - 0897914473

SN - 9780897914475

SP - 198

EP - 212

BT - Proceedings of the 13th ACM Symposium on Operating Systems Principles, SOSP 1991

ER -