VACED-SIM: A Simulator for Scalability Prediction in Large-Scale Parallel Computing

Yufei LIN; Xuejun YANG; Xinhai XU; Xiaowei GUO

doi:10.1587/transinf.E96.D.1430

VACED-SIM: A Simulator for Scalability Prediction in Large-Scale Parallel Computing

Yufei LIN, Xuejun YANG, Xinhai XU, Xiaowei GUO

Full Text Views

0

Share
Cite this

Summary :

Scaling up the system size has been the common approach to achieving high performance in parallel computing. However, designing and implementing a large-scale parallel system can be very costly in terms of money and time. When building a target system, it is desirable to initially build a smaller version by using the processing nodes with the same architecture as those in the target system. This allows us to achieve efficient and scalable prediction by using the smaller system to predict the performance of the target system. Such scalability prediction is critical because it enables system designers to evaluate different design alternatives so that a certain performance goal can be successfully achieved. As the de facto standard for writing parallel applications, MPI is widely used in large-scale parallel computing. By categorizing the discrete event simulation methods for MPI programs and analyzing the characteristics of scalability prediction, we propose a novel simulation method, called virtual-actual combined execution-driven (VACED) simulation, to achieve scalable prediction for MPI programs. The basic idea behind is to predict the execution time of an MPI program on a target machine by running it on a smaller system so that we can predict its communication time by virtual simulation and obtain its sequential computation time by actual execution. We introduce a model for the VACED simulation as well as the design and implementation of VACED-SIM, a lightweight simulator based on fine-grained activity and event definitions. We have validated our approach on a sub-system of Tianhe-1A. Our experimental results show that VACED-SIM exhibits higher accuracy and efficiency than MPI-SIM. In particular, for a target system with 1024 cores, the relative errors of VACED-SIM are less than 10% and the slowdowns are close to 1.

Publication: IEICE TRANSACTIONS on Information Vol.E96-D No.7 pp.1430-1442

Publication Date: 2013/07/01

Publicized

Online ISSN: 1745-1361

DOI: 10.1587/transinf.E96.D.1430

Type of Manuscript: PAPER

Category: Computer System

Authors

Yufei LIN
  National University of Defense Technology,National University of Defense Technology
Xuejun YANG
  National University of Defense Technology,National University of Defense Technology
Xinhai XU
  National University of Defense Technology,National University of Defense Technology
Xiaowei GUO
  National University of Defense Technology,National University of Defense Technology

Keyword

parallel computing, scalability prediction, MPI, communication primitive

Cite this

Copy

Yufei LIN, Xuejun YANG, Xinhai XU, Xiaowei GUO, "VACED-SIM: A Simulator for Scalability Prediction in Large-Scale Parallel Computing" in IEICE TRANSACTIONS on Information, vol. E96-D, no. 7, pp. 1430-1442, July 2013, doi: 10.1587/transinf.E96.D.1430.
Abstract: Scaling up the system size has been the common approach to achieving high performance in parallel computing. However, designing and implementing a large-scale parallel system can be very costly in terms of money and time. When building a target system, it is desirable to initially build a smaller version by using the processing nodes with the same architecture as those in the target system. This allows us to achieve efficient and scalable prediction by using the smaller system to predict the performance of the target system. Such scalability prediction is critical because it enables system designers to evaluate different design alternatives so that a certain performance goal can be successfully achieved. As the de facto standard for writing parallel applications, MPI is widely used in large-scale parallel computing. By categorizing the discrete event simulation methods for MPI programs and analyzing the characteristics of scalability prediction, we propose a novel simulation method, called virtual-actual combined execution-driven (VACED) simulation, to achieve scalable prediction for MPI programs. The basic idea behind is to predict the execution time of an MPI program on a target machine by running it on a smaller system so that we can predict its communication time by virtual simulation and obtain its sequential computation time by actual execution. We introduce a model for the VACED simulation as well as the design and implementation of VACED-SIM, a lightweight simulator based on fine-grained activity and event definitions. We have validated our approach on a sub-system of Tianhe-1A. Our experimental results show that VACED-SIM exhibits higher accuracy and efficiency than MPI-SIM. In particular, for a target system with 1024 cores, the relative errors of VACED-SIM are less than 10% and the slowdowns are close to 1.
URL: https://globals.ieice.org/en_transactions/information/10.1587/transinf.E96.D.1430/_p

Copy

@ARTICLE{e96-d_7_1430,
author={Yufei LIN, Xuejun YANG, Xinhai XU, Xiaowei GUO, },
journal={IEICE TRANSACTIONS on Information},
title={VACED-SIM: A Simulator for Scalability Prediction in Large-Scale Parallel Computing},
year={2013},
volume={E96-D},
number={7},
pages={1430-1442},
abstract={Scaling up the system size has been the common approach to achieving high performance in parallel computing. However, designing and implementing a large-scale parallel system can be very costly in terms of money and time. When building a target system, it is desirable to initially build a smaller version by using the processing nodes with the same architecture as those in the target system. This allows us to achieve efficient and scalable prediction by using the smaller system to predict the performance of the target system. Such scalability prediction is critical because it enables system designers to evaluate different design alternatives so that a certain performance goal can be successfully achieved. As the de facto standard for writing parallel applications, MPI is widely used in large-scale parallel computing. By categorizing the discrete event simulation methods for MPI programs and analyzing the characteristics of scalability prediction, we propose a novel simulation method, called virtual-actual combined execution-driven (VACED) simulation, to achieve scalable prediction for MPI programs. The basic idea behind is to predict the execution time of an MPI program on a target machine by running it on a smaller system so that we can predict its communication time by virtual simulation and obtain its sequential computation time by actual execution. We introduce a model for the VACED simulation as well as the design and implementation of VACED-SIM, a lightweight simulator based on fine-grained activity and event definitions. We have validated our approach on a sub-system of Tianhe-1A. Our experimental results show that VACED-SIM exhibits higher accuracy and efficiency than MPI-SIM. In particular, for a target system with 1024 cores, the relative errors of VACED-SIM are less than 10% and the slowdowns are close to 1.},
keywords={},
doi={10.1587/transinf.E96.D.1430},
ISSN={1745-1361},
month={July},}

Copy

TY - JOUR
TI - VACED-SIM: A Simulator for Scalability Prediction in Large-Scale Parallel Computing
T2 - IEICE TRANSACTIONS on Information
SP - 1430
EP - 1442
AU - Yufei LIN
AU - Xuejun YANG
AU - Xinhai XU
AU - Xiaowei GUO
PY - 2013
DO - 10.1587/transinf.E96.D.1430
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E96-D
IS - 7
JA - IEICE TRANSACTIONS on Information
Y1 - July 2013
AB - Scaling up the system size has been the common approach to achieving high performance in parallel computing. However, designing and implementing a large-scale parallel system can be very costly in terms of money and time. When building a target system, it is desirable to initially build a smaller version by using the processing nodes with the same architecture as those in the target system. This allows us to achieve efficient and scalable prediction by using the smaller system to predict the performance of the target system. Such scalability prediction is critical because it enables system designers to evaluate different design alternatives so that a certain performance goal can be successfully achieved. As the de facto standard for writing parallel applications, MPI is widely used in large-scale parallel computing. By categorizing the discrete event simulation methods for MPI programs and analyzing the characteristics of scalability prediction, we propose a novel simulation method, called virtual-actual combined execution-driven (VACED) simulation, to achieve scalable prediction for MPI programs. The basic idea behind is to predict the execution time of an MPI program on a target machine by running it on a smaller system so that we can predict its communication time by virtual simulation and obtain its sequential computation time by actual execution. We introduce a model for the VACED simulation as well as the design and implementation of VACED-SIM, a lightweight simulator based on fine-grained activity and event definitions. We have validated our approach on a sub-system of Tianhe-1A. Our experimental results show that VACED-SIM exhibits higher accuracy and efficiency than MPI-SIM. In particular, for a target system with 1024 cores, the relative errors of VACED-SIM are less than 10% and the slowdowns are close to 1.
ER -