Monetary cost optimizations for MPI-based HPC applications on Amazon clouds: Checkpoints and replicated execution

Yifan Gong; Bingsheng He; Amelie Chi Zhou

doi:10.1145/2807591.2807612

Monetary cost optimizations for MPI-based HPC applications on Amazon clouds: Checkpoints and replicated execution

Yifan Gong
, Bingsheng He
, Amelie Chi Zhou

Department of Computer Science

Research output: Chapter in book/report/conference proceeding › Conference proceeding › peer-review

25 Citations (Scopus)

Abstract

In this paper, we propose monetary cost optimizations for MPI-based applications with deadline constraints on Amazon EC2. Particularly, we consider to utilize two kinds of Amazon EC2 instances (on-demand and spot instances). As a spot instance can fail at any time due to out-of-bid events, fault tolerant executions are necessary. Through detailed studies, we have found that two common fault tolerant mechanisms, i.e., checkpoints and replicated executions, are complementary for cost-effective MPI executions on spot instances. We formulate the optimization problem and propose a novel cost model to minimize the expected monetary cost. The experimental results with NPB benchmarks on Amazon EC2 demonstrate that 1) it is feasible to run MPI applications with performance constraints on spot instances, 2) our proposal achieves significant monetary cost reduction compared to the state-of-the-art algorithm and 3) it is necessary to adaptively choose checkpoint and replication techniques for cost-effective and reliable MPI executions on Amazon EC2.

Original language	English
Title of host publication	SC 2015: Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis
Place of Publication	New York
Publisher	IEEE
Number of pages	12
ISBN (Electronic)	9781450337236
DOIs	https://doi.org/10.1145/2807591.2807612
Publication status	Published - 15 Nov 2015
Event	International Conference for High Performance Computing, Networking, Storage and Analysis - Austin, United States Duration: 15 Nov 2015 → 20 Nov 2015 https://sc15.supercomputing.org/ (Conference website) https://dl.acm.org/doi/proceedings/10.1145/2807591 (Conference proceeding)

Publication series

Name	International Conference for High Performance Computing, Networking, Storage and Analysis, SC
Volume	November-2015
ISSN (Print)	2167-4329
ISSN (Electronic)	2167-4337

Conference

Conference	International Conference for High Performance Computing, Networking, Storage and Analysis
Abbreviated title	SC 2015
Country/Territory	United States
City	Austin
Period	15/11/15 → 20/11/15
Internet address	https://sc15.supercomputing.org/ (Conference website) https://dl.acm.org/doi/proceedings/10.1145/2807591 (Conference proceeding)

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 9 Industry, Innovation, and Infrastructure

User-Defined Keywords

cloud
fault-tolerance
monetary cost
MPI
spot instance

Access to Document

10.1145/2807591.2807612

Cite this

Gong, Y., He, B., & Zhou, A. C. (2015). Monetary cost optimizations for MPI-based HPC applications on Amazon clouds: Checkpoints and replicated execution. In SC 2015: Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis Article 32 (International Conference for High Performance Computing, Networking, Storage and Analysis, SC; Vol. November-2015). IEEE. https://doi.org/10.1145/2807591.2807612

Gong, Yifan ; He, Bingsheng ; Zhou, Amelie Chi. / Monetary cost optimizations for MPI-based HPC applications on Amazon clouds : Checkpoints and replicated execution. SC 2015: Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis. New York : IEEE, 2015. (International Conference for High Performance Computing, Networking, Storage and Analysis, SC).

@inproceedings{73b3b64eb6664ed8b761800c905d716c,

title = "Monetary cost optimizations for MPI-based HPC applications on Amazon clouds: Checkpoints and replicated execution",

abstract = "In this paper, we propose monetary cost optimizations for MPI-based applications with deadline constraints on Amazon EC2. Particularly, we consider to utilize two kinds of Amazon EC2 instances (on-demand and spot instances). As a spot instance can fail at any time due to out-of-bid events, fault tolerant executions are necessary. Through detailed studies, we have found that two common fault tolerant mechanisms, i.e., checkpoints and replicated executions, are complementary for cost-effective MPI executions on spot instances. We formulate the optimization problem and propose a novel cost model to minimize the expected monetary cost. The experimental results with NPB benchmarks on Amazon EC2 demonstrate that 1) it is feasible to run MPI applications with performance constraints on spot instances, 2) our proposal achieves significant monetary cost reduction compared to the state-of-the-art algorithm and 3) it is necessary to adaptively choose checkpoint and replication techniques for cost-effective and reliable MPI executions on Amazon EC2.",

keywords = "cloud, fault-tolerance, monetary cost, MPI, spot instance",

author = "Yifan Gong and Bingsheng He and Zhou, \{Amelie Chi\}",

note = "We acknowledge the support from the Singapore National Research Foundation under its Environmental \& Water Technologies Strategic Research Programme and administered by the Environment \& Water Industry Programme Office (EWI) of the PUB, under project 1002-IRIS-09. Yifan and Amelie are supported by the scholarship from NEWRI, IGS(Interdisciplinary Graduate School). Bingsheng is in part supported by a MoE AcRF Tier 1 (2014-T1-001-145) of Singapore. Publisher Copyright: {\textcopyright} 2015 ACM.; International Conference for High Performance Computing, Networking, Storage and Analysis, SC 2015 ; Conference date: 15-11-2015 Through 20-11-2015",

year = "2015",

month = nov,

day = "15",

doi = "10.1145/2807591.2807612",

language = "English",

series = "International Conference for High Performance Computing, Networking, Storage and Analysis, SC",

publisher = "IEEE",

booktitle = "SC 2015: Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis",

address = "United States",

url = "https://sc15.supercomputing.org/, https://dl.acm.org/doi/proceedings/10.1145/2807591",

}

Gong, Y, He, B & Zhou, AC 2015, Monetary cost optimizations for MPI-based HPC applications on Amazon clouds: Checkpoints and replicated execution. in SC 2015: Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis., 32, International Conference for High Performance Computing, Networking, Storage and Analysis, SC, vol. November-2015, IEEE, New York, International Conference for High Performance Computing, Networking, Storage and Analysis, Austin, Texas, United States, 15/11/15. https://doi.org/10.1145/2807591.2807612

Monetary cost optimizations for MPI-based HPC applications on Amazon clouds: Checkpoints and replicated execution. / Gong, Yifan; He, Bingsheng; Zhou, Amelie Chi.
SC 2015: Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis. New York: IEEE, 2015. 32 (International Conference for High Performance Computing, Networking, Storage and Analysis, SC; Vol. November-2015).

Research output: Chapter in book/report/conference proceeding › Conference proceeding › peer-review

TY - GEN

T1 - Monetary cost optimizations for MPI-based HPC applications on Amazon clouds

T2 - International Conference for High Performance Computing, Networking, Storage and Analysis

AU - Gong, Yifan

AU - He, Bingsheng

AU - Zhou, Amelie Chi

N1 - We acknowledge the support from the Singapore National Research Foundation under its Environmental & Water Technologies Strategic Research Programme and administered by the Environment & Water Industry Programme Office (EWI) of the PUB, under project 1002-IRIS-09. Yifan and Amelie are supported by the scholarship from NEWRI, IGS(Interdisciplinary Graduate School). Bingsheng is in part supported by a MoE AcRF Tier 1 (2014-T1-001-145) of Singapore. Publisher Copyright: © 2015 ACM.

PY - 2015/11/15

Y1 - 2015/11/15

N2 - In this paper, we propose monetary cost optimizations for MPI-based applications with deadline constraints on Amazon EC2. Particularly, we consider to utilize two kinds of Amazon EC2 instances (on-demand and spot instances). As a spot instance can fail at any time due to out-of-bid events, fault tolerant executions are necessary. Through detailed studies, we have found that two common fault tolerant mechanisms, i.e., checkpoints and replicated executions, are complementary for cost-effective MPI executions on spot instances. We formulate the optimization problem and propose a novel cost model to minimize the expected monetary cost. The experimental results with NPB benchmarks on Amazon EC2 demonstrate that 1) it is feasible to run MPI applications with performance constraints on spot instances, 2) our proposal achieves significant monetary cost reduction compared to the state-of-the-art algorithm and 3) it is necessary to adaptively choose checkpoint and replication techniques for cost-effective and reliable MPI executions on Amazon EC2.

AB - In this paper, we propose monetary cost optimizations for MPI-based applications with deadline constraints on Amazon EC2. Particularly, we consider to utilize two kinds of Amazon EC2 instances (on-demand and spot instances). As a spot instance can fail at any time due to out-of-bid events, fault tolerant executions are necessary. Through detailed studies, we have found that two common fault tolerant mechanisms, i.e., checkpoints and replicated executions, are complementary for cost-effective MPI executions on spot instances. We formulate the optimization problem and propose a novel cost model to minimize the expected monetary cost. The experimental results with NPB benchmarks on Amazon EC2 demonstrate that 1) it is feasible to run MPI applications with performance constraints on spot instances, 2) our proposal achieves significant monetary cost reduction compared to the state-of-the-art algorithm and 3) it is necessary to adaptively choose checkpoint and replication techniques for cost-effective and reliable MPI executions on Amazon EC2.

KW - cloud

KW - fault-tolerance

KW - monetary cost

KW - MPI

KW - spot instance

UR - https://www.scopus.com/pages/publications/84966656934

UR - https://dl.acm.org/doi/10.1145/2807591.2807612

U2 - 10.1145/2807591.2807612

DO - 10.1145/2807591.2807612

M3 - Conference proceeding

AN - SCOPUS:84966656934

T3 - International Conference for High Performance Computing, Networking, Storage and Analysis, SC

BT - SC 2015: Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis

PB - IEEE

CY - New York

Y2 - 15 November 2015 through 20 November 2015

ER -

Gong Y, He B, Zhou AC. Monetary cost optimizations for MPI-based HPC applications on Amazon clouds: Checkpoints and replicated execution. In SC 2015: Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis. New York: IEEE. 2015. 32. (International Conference for High Performance Computing, Networking, Storage and Analysis, SC). doi: 10.1145/2807591.2807612

Monetary cost optimizations for MPI-based HPC applications on Amazon clouds: Checkpoints and replicated execution

Abstract

Publication series

Conference

UN SDGs

User-Defined Keywords

Access to Document

Other files and links

Fingerprint

Cite this