Dragonfly: A multi-platform parallel toolbox for MATLAB/Octave

被引:2
作者
Azzini, Ivano [2 ]
Muresano, Ronal [1 ]
Ratto, Marco [1 ]
机构
[1] European Commiss, Joint Res Ctr, Ispra, Italy
[2] GET Italia Srl, Milan, Italy
关键词
Parallel tool; Efficiency; High-performance computing (HPC); Performance metrics; MATLAB and Octave executions; Multi-platform; MATLAB;
D O I
10.1016/j.cl.2017.10.002
中图分类号
TP31 [计算机软件];
学科分类号
081202 ; 0835 ;
摘要
In this paper, we present Dragonfly, a free open-source software toolkit that allows creating a parallel collaboration environment between MATLAB and/or Octave instances, allowing us to have a robust system, which is flexible, transparent and interoperable. Dragonfly can be used under multi-platform (Linux, Windows, Mac) and multi-environment configurations (e.g., MATLAB and/or Octave instances). To distribute the workload, it has a flexible system, where programmers can select the computational resources, create clusters and assign weights to balance properly tasks across the distributed environment (MATLAB and Octave instances machines) with different computational power. These functionalities allow obtaining good performance metrics such as: execution time, scalability and efficiency. Finally, this tool has been tested on different computer systems from standard commodity environments to medium size clusters using a set of applications from different scientific fields. This widespread in-field validation allowed showing good performances not only in speedup but also in efficiency using Dragonfly. (C) 2017 The Authors. Published by Elsevier Ltd.
引用
收藏
页码:21 / 42
页数:22
相关论文
共 25 条
[1]  
Abdi Z., 2014, ARXIV14076878
[2]  
Adjemian S., 2011, DYNARE WORKING PAPER
[3]  
[Anonymous], 1995, Designing and Building Parallel Programs
[4]  
Azzini I, 2007, P 3 DYNARE C FRANC P
[5]  
Azzini I, 2011, P 17 INT C COMP EC F
[6]  
Barney B., 2012, Introduction to parallel computing
[7]  
Bliss N.T., 2007, 2007 IEEE INT C AC S, V4, pIV 1189, DOI [10.1109/ICASSP.2007.367288, DOI 10.1109/ICASSP.2007.367288]
[8]   pMATLAB parallel MATLAB library [J].
Bliss, N. Travinin ;
Kepner, J. .
INTERNATIONAL JOURNAL OF HIGH PERFORMANCE COMPUTING APPLICATIONS, 2007, 21 (03) :336-359
[9]  
Buyya R., 1999, HIGH PERFORMANCE CLU, V1
[10]   Parallel MATLAB: Doing it right [J].
Choy, R ;
Edelman, A .
PROCEEDINGS OF THE IEEE, 2005, 93 (02) :331-341