Managing performance analysis with dynamic statistical projection pursuit

Jeffrey S. Vetter; Daniel A. Reed

doi:10.1109/SC.1999.10028

Managing performance analysis with dynamic statistical projection pursuit

Jeffrey S. Vetter, Daniel A. Reed

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

15 Scopus citations

Abstract

Computer systems and applications are growing more complex. Consequently, performance analysis has become more difficult due to the complex, transient interrelationships among runtime components. To diagnose these types of performance issues, developers must use detailed instrumentation to capture a large number of performance metrics. Unfortunately, this instrumentation may actually influence the performance analysis, leading the developer to an ambiguous conclusion. In this paper, we introduce a technique for focussing a performance analysis on interesting performance metrics. This technique, called dynamic statistical projection pursuit, identifies interesting performance metrics that the monitoring system should capture across some number of processors. By reducing the number of performance metrics, projection pursuit can limit the impact of instrumentation on the performance of the target system and can reduce the volume of performance data.

Original language	English
Title of host publication	ACM/IEEE SC 1999 Conference, SC 1999
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	44
Number of pages	1
ISBN (Electronic)	1581130910, 9781581130911
DOIs	https://doi.org/10.1109/SC.1999.10028
State	Published - 1999
Externally published	Yes
Event	1999 ACM/IEEE Conference on Supercomputing, SC 1999 - Portland, United States Duration: Nov 13 1999 → Nov 19 1999

Publication series

Name	ACM/IEEE SC 1999 Conference, SC 1999

Conference

Conference	1999 ACM/IEEE Conference on Supercomputing, SC 1999
Country/Territory	United States
City	Portland
Period	11/13/99 → 11/19/99

Funding

1 This work was supported in part by the Defense Advanced Research Projects Agency under DARPA contracts DABT63-94-C0049 (SIO Initiative), F30602-96-C-0161, and DABT63-96-C-0027 by the National Science Foundation under grants NSF CDA 94-01124 and ASC 97-20202, and by the Department of Energy under contracts DOE B-341494, W-7405-ENG-48, and 1-B-333164. Appears in Proc. SC99, Portland, Oregon, USA (November 1999). This work was supported in part by the Defense Advanced Research Projects Agency under DARPA contracts DABT63-94-C0049 (SIO Initiative), F30602-96-C-0161, and DABT63-96-C-0027 by the National Science Foundation under grants NSF CDA 94-01124 and ASC 97-20202, and by the Department of Energy under contracts DOE B-341494, W-7405-ENG-48, and 1-B-333164. This paper has benefited from the detailed comments of our SC99 reviewers and our colleagues in the Pablo group at the University of Illinois at Urbana-Champaign and at Lawrence Livermore National Laboratory. We are also grateful to the DOE ASCI (http://www.llnl.gov/ASCI) program for making the ASCI benchmarks available and to the Cactus team (http://www.cactuscode.org) for providing us with the Cactus framework.

Funders	Funder number
National Science Foundation	ASC 97-20202, CDA 94-01124
U.S. Department of Energy	1-B-333164, W-7405-ENG-48, DOE B-341494
Defense Advanced Research Projects Agency	F30602-96-C-0161, DABT63-96-C-0027, DABT63-94-C0049
National Science Foundation

Access to Document

10.1109/SC.1999.10028

Cite this

@inproceedings{dab89cb9c96b4680a22b3b9643594c52,

title = "Managing performance analysis with dynamic statistical projection pursuit",

abstract = "Computer systems and applications are growing more complex. Consequently, performance analysis has become more difficult due to the complex, transient interrelationships among runtime components. To diagnose these types of performance issues, developers must use detailed instrumentation to capture a large number of performance metrics. Unfortunately, this instrumentation may actually influence the performance analysis, leading the developer to an ambiguous conclusion. In this paper, we introduce a technique for focussing a performance analysis on interesting performance metrics. This technique, called dynamic statistical projection pursuit, identifies interesting performance metrics that the monitoring system should capture across some number of processors. By reducing the number of performance metrics, projection pursuit can limit the impact of instrumentation on the performance of the target system and can reduce the volume of performance data.",

author = "Vetter, {Jeffrey S.} and Reed, {Daniel A.}",

note = "Publisher Copyright: {\textcopyright} 1999 IEEE.; 1999 ACM/IEEE Conference on Supercomputing, SC 1999 ; Conference date: 13-11-1999 Through 19-11-1999",

year = "1999",

doi = "10.1109/SC.1999.10028",

language = "English",

series = "ACM/IEEE SC 1999 Conference, SC 1999",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "44",

booktitle = "ACM/IEEE SC 1999 Conference, SC 1999",

}

Vetter, JS & Reed, DA 1999, Managing performance analysis with dynamic statistical projection pursuit. in ACM/IEEE SC 1999 Conference, SC 1999., 1592687, ACM/IEEE SC 1999 Conference, SC 1999, Institute of Electrical and Electronics Engineers Inc., pp. 44, 1999 ACM/IEEE Conference on Supercomputing, SC 1999, Portland, United States, 11/13/99. https://doi.org/10.1109/SC.1999.10028

Managing performance analysis with dynamic statistical projection pursuit. / Vetter, Jeffrey S.; Reed, Daniel A.
ACM/IEEE SC 1999 Conference, SC 1999. Institute of Electrical and Electronics Engineers Inc., 1999. p. 44 1592687 (ACM/IEEE SC 1999 Conference, SC 1999).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Managing performance analysis with dynamic statistical projection pursuit

AU - Vetter, Jeffrey S.

AU - Reed, Daniel A.

PY - 1999

Y1 - 1999

N2 - Computer systems and applications are growing more complex. Consequently, performance analysis has become more difficult due to the complex, transient interrelationships among runtime components. To diagnose these types of performance issues, developers must use detailed instrumentation to capture a large number of performance metrics. Unfortunately, this instrumentation may actually influence the performance analysis, leading the developer to an ambiguous conclusion. In this paper, we introduce a technique for focussing a performance analysis on interesting performance metrics. This technique, called dynamic statistical projection pursuit, identifies interesting performance metrics that the monitoring system should capture across some number of processors. By reducing the number of performance metrics, projection pursuit can limit the impact of instrumentation on the performance of the target system and can reduce the volume of performance data.

AB - Computer systems and applications are growing more complex. Consequently, performance analysis has become more difficult due to the complex, transient interrelationships among runtime components. To diagnose these types of performance issues, developers must use detailed instrumentation to capture a large number of performance metrics. Unfortunately, this instrumentation may actually influence the performance analysis, leading the developer to an ambiguous conclusion. In this paper, we introduce a technique for focussing a performance analysis on interesting performance metrics. This technique, called dynamic statistical projection pursuit, identifies interesting performance metrics that the monitoring system should capture across some number of processors. By reducing the number of performance metrics, projection pursuit can limit the impact of instrumentation on the performance of the target system and can reduce the volume of performance data.

UR - http://www.scopus.com/inward/record.url?scp=84905475508&partnerID=8YFLogxK

U2 - 10.1109/SC.1999.10028

DO - 10.1109/SC.1999.10028

M3 - Conference contribution

AN - SCOPUS:84905475508

T3 - ACM/IEEE SC 1999 Conference, SC 1999

SP - 44

BT - ACM/IEEE SC 1999 Conference, SC 1999

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 1999 ACM/IEEE Conference on Supercomputing, SC 1999

Y2 - 13 November 1999 through 19 November 1999

ER -

Managing performance analysis with dynamic statistical projection pursuit

Abstract

Publication series

Conference

Funding

Access to Document

Other files and links

Fingerprint

Cite this