Group-Based Performance Analysis for Multithreaded SMP Cluster Applications

Brunst, Holger; Nagel, Wolfgang E.; Hoppe, Hans–Christian

doi:10.1007/3-540-44681-8_22

Cited by 10 publications

(4 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The performance evaluation and the subsequent optimization of a parallel program executed on a GRID imposes special requirements on the generation and visualization of performance data [4]. In order to allow for a largely transparent view of the application's behavior, there is a need for some background mechanism that collects and merges the performance data generated on the distributed nodes of the GRID.…”

Section: How To Adapt Existing Analysis Technologymentioning

confidence: 99%

Performance Analysis of a Parallel Application in the GRID

Brunst¹,

Gabriel

Lange

et al. 2003

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

Performance analysis of real applications in clusters and GRID like environments is essential to fully exploit the performance of new architectures. The key problem is the deepening hierarchy of latencies and bandwidths and the growing heterogeneity of systems. This paper discusses the basic problems of performance analysis in such clustered and heterogeneous environments. It further presents a software environment that supports the user in running codes and getting more insight into the performance of the application. In order to give a proof of the concept a code for direct numerical simulation of reactive flowsi sr u ni naGRID like hardware environment, and the performance analysis is presented.

show abstract

Section: How To Adapt Existing Analysis Technologymentioning

confidence: 99%

Performance Analysis of a Parallel Application in the GRID

Brunst¹,

Gabriel

Lange

et al. 2003

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

show abstract

“…Experiments were conducted to evaluate the trace processing improvements when using multiple cluster nodes for analysis. These results are shown for two different platforms and compared to baseline performance for the Vampir [3,4] trace visualization system. The paper concludes with a discussion of the work and directions for the future.…”

Section: Introductionmentioning

confidence: 99%

A distributed performance analysis architecture for clusters

Brunst¹,

Nagel²,

Malony³

2003

Proceedings IEEE International Conference on Cluster Computing CLUSTR-03

View full text Add to dashboard Cite

show abstract

“…The new approach was motivated by the work [2,3,4] about analysis and visualization of very large and highly parallel program traces of the local research group -especially in conjunction with the adaption of VAMPIR towards the ASCI project.…”

Section: Introductionmentioning

confidence: 99%

A New Data Compression Technique for Event Based Program Traces

Knüpfer

2003

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. The paper presents an innovative solution to the problem of the very huge data sets that are regularly produced by performance tracing techniques -especially on HPC programs. It designs an adapted data compression scheme that takes advantage of regularities frequently found in program traces. Algorithms to reveal repetition patterns in a programs call structure and run time behavior are discussed in detail, solutions to some problems arising on practical application are addressed as well. Two examples demonstrate the capabilities of the approach and document its behavior. Finally, some thoughts are given regarding how the patterns revealed in the process of data compression may assist the automatic analysis of traces.

show abstract

Group-Based Performance Analysis for Multithreaded SMP Cluster Applications

Cited by 10 publications

References 3 publications

Performance Analysis of a Parallel Application in the GRID

Performance Analysis of a Parallel Application in the GRID

A distributed performance analysis architecture for clusters

A New Data Compression Technique for Event Based Program Traces

Contact Info

Product

Resources

About