A multi-level monitoring framework for stream-based coordination programs

Vu Thien Nga Nguyen, Raimund Kirner, Frank Penczek

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

3 Citations (Scopus)

Abstract

Stream-based Coordination is a promising approach to execute programs on parallel hardware such as multi-core systems. It allows to reuse sequential code at component level and to extend such code with concurrency-handling at the coordination level. In this paper we identify the monitoring information required to enable the calculation of performance metrics, automatic load balancing, and bottleneck detection. The monitoring information is obtained by implicitly instrumenting multiple levels: the runtime system and the operating system. We evaluate the monitoring overhead caused by different use cases on S-Net as it is a challenging monitoring benchmark with a flexible and fully asynchronous execution model, including dynamic mapping and scheduling policies. The evaluation shows that in most cases the monitoring causes a negligible overhead of less than five percent.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages83-98
Number of pages16
Volume7439 LNCS
EditionPART 1
DOIs
Publication statusPublished - 2012
Event12th International Conference on Algorithms and Architectures for Parallel Processing, ICA3PP 2012 - Fukuoka, Japan
Duration: 4 Sept 20127 Sept 2012

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 1
Volume7439 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Conference

Conference12th International Conference on Algorithms and Architectures for Parallel Processing, ICA3PP 2012
Country/TerritoryJapan
CityFukuoka
Period4/09/127/09/12

Fingerprint

Dive into the research topics of 'A multi-level monitoring framework for stream-based coordination programs'. Together they form a unique fingerprint.

Cite this