Research Article

Improving the Design of Parallel Applications using Statistical Methods

C.M. Amarandei, D. Lepadatu and S. Caraiman

The aim of this study is to analyze data produced by a parallel application using statistical investigation methods like design of experiments to determine faults in the design or implementation of the application. Using this technique we develop a model to verify the design of a parallel application by determining the factors influencing its performance and describe the behavior of the application based on these factors. Experimental studies were performed on a parallel rendering application based on a sort-last domain decomposition technique. The following characteristics of the parallel application were analyzed: processing time, communication time and size of the data exchanged between processes. The performed analysis revealed the limitations of the application design in various runtime conditions offering strong clues on how certain stages of the application should be redesigned.

PDF Fulltext XML References Citation

How to cite this article

C.M. Amarandei, D. Lepadatu and S. Caraiman, 2011. Improving the Design of Parallel Applications using Statistical Methods. Journal of Applied Sciences, 11: 932-942.

DOI: 10.3923/jas.2011.932.942

URL: https://scialert.net/abstract/?doi=jas.2011.932.942

Journal of Applied Sciences

Article Trend

Total views 577

Search

Authors

C.M. Amarandei

D. Lepadatu

S. Caraiman

Keywords

Research Article

Improving the Design of Parallel Applications using Statistical Methods

How to cite this article

Related Articles

Leave a Comment