IEICE TRANSACTIONS on Information
Entropy Based Evaluation of Communication Predictability in Parallel Applications
Summary :
The performance of parallel computing applications is highly dependent on the efficiency of the underlying communication operations. While often characterized as dynamic, these communication operations frequently exhibit spatial and temporal locality as well as regularity in structure. These characteristics can be exploited to improve communication performance if the correct prediction model is selected to a suitable communication topology. In this paper we describe an entropy based methodology for quantifying and evaluating the success of different prediction models on actual workloads drawn from representative parallel benchmarks. We evaluate two different prediction criteria and combinations thereof: (1) Messages are partitioned by source node. (2) Use of a first order context model. We also describe the threshold for predication designed to largely avoid incorrect predication overheads. Our results show for simple predication models, even on highly dynamic benchmark applications, predictability can be improved by several orders of magnitude. In fact, using simple prediction techniques, over 75% of the communication volume is accurately predictable.
- Publication
- IEICE TRANSACTIONS on Information Vol.E89-D No.2 pp.469-478
- Publication Date
- 2006/02/01
- Publicized
- Online ISSN
- 1745-1361
- DOI
- 10.1093/ietisy/e89-d.2.469
- Type of Manuscript
- Special Section PAPER (Special Section on Parallel/Distributed Computing and Networking)
- Category
- Performance Evaluation
Authors
Keyword
Latest Issue
Copyrights notice of machine-translated contents
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Cite this
Copy
Alex K. JONES, Jiang ZHENG, Ahmed AMER, "Entropy Based Evaluation of Communication Predictability in Parallel Applications" in IEICE TRANSACTIONS on Information,
vol. E89-D, no. 2, pp. 469-478, February 2006, doi: 10.1093/ietisy/e89-d.2.469.
Abstract: The performance of parallel computing applications is highly dependent on the efficiency of the underlying communication operations. While often characterized as dynamic, these communication operations frequently exhibit spatial and temporal locality as well as regularity in structure. These characteristics can be exploited to improve communication performance if the correct prediction model is selected to a suitable communication topology. In this paper we describe an entropy based methodology for quantifying and evaluating the success of different prediction models on actual workloads drawn from representative parallel benchmarks. We evaluate two different prediction criteria and combinations thereof: (1) Messages are partitioned by source node. (2) Use of a first order context model. We also describe the threshold for predication designed to largely avoid incorrect predication overheads. Our results show for simple predication models, even on highly dynamic benchmark applications, predictability can be improved by several orders of magnitude. In fact, using simple prediction techniques, over 75% of the communication volume is accurately predictable.
URL: https://global.ieice.org/en_transactions/information/10.1093/ietisy/e89-d.2.469/_p
Copy
@ARTICLE{e89-d_2_469,
author={Alex K. JONES, Jiang ZHENG, Ahmed AMER, },
journal={IEICE TRANSACTIONS on Information},
title={Entropy Based Evaluation of Communication Predictability in Parallel Applications},
year={2006},
volume={E89-D},
number={2},
pages={469-478},
abstract={The performance of parallel computing applications is highly dependent on the efficiency of the underlying communication operations. While often characterized as dynamic, these communication operations frequently exhibit spatial and temporal locality as well as regularity in structure. These characteristics can be exploited to improve communication performance if the correct prediction model is selected to a suitable communication topology. In this paper we describe an entropy based methodology for quantifying and evaluating the success of different prediction models on actual workloads drawn from representative parallel benchmarks. We evaluate two different prediction criteria and combinations thereof: (1) Messages are partitioned by source node. (2) Use of a first order context model. We also describe the threshold for predication designed to largely avoid incorrect predication overheads. Our results show for simple predication models, even on highly dynamic benchmark applications, predictability can be improved by several orders of magnitude. In fact, using simple prediction techniques, over 75% of the communication volume is accurately predictable.},
keywords={},
doi={10.1093/ietisy/e89-d.2.469},
ISSN={1745-1361},
month={February},}
Copy
TY - JOUR
TI - Entropy Based Evaluation of Communication Predictability in Parallel Applications
T2 - IEICE TRANSACTIONS on Information
SP - 469
EP - 478
AU - Alex K. JONES
AU - Jiang ZHENG
AU - Ahmed AMER
PY - 2006
DO - 10.1093/ietisy/e89-d.2.469
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E89-D
IS - 2
JA - IEICE TRANSACTIONS on Information
Y1 - February 2006
AB - The performance of parallel computing applications is highly dependent on the efficiency of the underlying communication operations. While often characterized as dynamic, these communication operations frequently exhibit spatial and temporal locality as well as regularity in structure. These characteristics can be exploited to improve communication performance if the correct prediction model is selected to a suitable communication topology. In this paper we describe an entropy based methodology for quantifying and evaluating the success of different prediction models on actual workloads drawn from representative parallel benchmarks. We evaluate two different prediction criteria and combinations thereof: (1) Messages are partitioned by source node. (2) Use of a first order context model. We also describe the threshold for predication designed to largely avoid incorrect predication overheads. Our results show for simple predication models, even on highly dynamic benchmark applications, predictability can be improved by several orders of magnitude. In fact, using simple prediction techniques, over 75% of the communication volume is accurately predictable.
ER -
English
