TimeTrial: A low-impact performance profiler for streaming data applications

Joseph M. Lancaster, E. F.Berkley Shands, Jeremy D. Buhler, Roger D. Chamberlain

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

5 Scopus citations

Abstract

Finding performance bottlenecks in application-specific systems is becoming increasingly labor-intensive as the capabilities of these systems improve. The complex platforms required to meet today's high application performance demands put pressure on developers to sustain current design cycles. Application developers need better tools to diagnose performance issues that arise when utilizing real-world application-specific platforms, from embedded applications to high-performance computational science applications. In this paper, we present TimeTrial, a runtime performance monitoring system that profiles streaming data applications deployed on architecturally diverse computers. TimeTrial is designed to operate with minimal impact on the executing application, exploiting user directives to aggressively perform lossy compression on performance meta-data. We present the design of the TimeTrial performance monitor and demonstrate its use in discovering performance bottlenecks in a production computational biology application.

Original languageEnglish
Title of host publicationProceedings - 22nd IEEE International Conference on Application-Specific Systems, Architectures and Processors, ASAP 2011
Pages69-76
Number of pages8
DOIs
StatePublished - 2011
Event22nd IEEE International Conference on Application-Specific Systems, Architectures and Processors, ASAP 2011 - Santa Monica, CA, United States
Duration: Sep 11 2011Sep 14 2011

Publication series

NameProceedings of the International Conference on Application-Specific Systems, Architectures and Processors
ISSN (Print)1063-6862

Conference

Conference22nd IEEE International Conference on Application-Specific Systems, Architectures and Processors, ASAP 2011
Country/TerritoryUnited States
CitySanta Monica, CA
Period09/11/1109/14/11

Keywords

  • Runtime Monitor
  • Stream Computing

Fingerprint

Dive into the research topics of 'TimeTrial: A low-impact performance profiler for streaming data applications'. Together they form a unique fingerprint.

Cite this