Accelerating read atomic multi-partition transaction with remote direct memory access

Naofumi Murata, Hideyuki Kawashima, Osamu Tatebe

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Many applications these days require data processing that is both efficient and reliable. Distributed databases are one way to meet these requirements, but must be updated using distributed transactions. To manage foreign key constraints, secondary indices, and materialized views in distributed environments, read atomic multi-partition (RAMP) transactions demonstrate high efficiency. RAMP transactions achieved high performance distributed transaction processing by relaxing the isolation level. However, the use of fast interconnect to accelerate performance has not yet been considered. This paper proposes the acceleration of RAMP transactions by exploiting remote direct memory access (RDMA) on the InfiniBand interconnect. We first present GET+ and PUT+ operations that accelerate the RAMP transaction by exploiting RDMA write operations. We then present the GET∗ operation, which further accelerates GET+ operations exploiting RDMA read operations. To evaluate the proposed methods, we implemented a prototype distributed in-memory key-value store in C/C++. The results of the experiments show that compared with RAMP transactions on IP over InfiniBand, GET∗ and PUT+ achieve a 2.67× performance improvement on the Yahoo! Cloud Serving Benchmark. All of our code is publicly available.

Original languageEnglish
Title of host publication2017 IEEE International Conference on Big Data and Smart Computing, BigComp 2017
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages239-246
Number of pages8
ISBN (Electronic)9781509030156
DOIs
Publication statusPublished - 2017 Mar 17
Externally publishedYes
Event2017 IEEE International Conference on Big Data and Smart Computing, BigComp 2017 - Jeju Island, Korea, Republic of
Duration: 2017 Feb 132017 Feb 16

Other

Other2017 IEEE International Conference on Big Data and Smart Computing, BigComp 2017
CountryKorea, Republic of
CityJeju Island
Period17/2/1317/2/16

Fingerprint

Data storage equipment
Processing
Experiments

Keywords

  • Distributed Transaction
  • RAMP Transaction
  • RDMA

ASJC Scopus subject areas

  • Information Systems
  • Artificial Intelligence
  • Computer Science Applications
  • Computer Vision and Pattern Recognition

Cite this

Murata, N., Kawashima, H., & Tatebe, O. (2017). Accelerating read atomic multi-partition transaction with remote direct memory access. In 2017 IEEE International Conference on Big Data and Smart Computing, BigComp 2017 (pp. 239-246). [7881705] Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/BIGCOMP.2017.7881705

Accelerating read atomic multi-partition transaction with remote direct memory access. / Murata, Naofumi; Kawashima, Hideyuki; Tatebe, Osamu.

2017 IEEE International Conference on Big Data and Smart Computing, BigComp 2017. Institute of Electrical and Electronics Engineers Inc., 2017. p. 239-246 7881705.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Murata, N, Kawashima, H & Tatebe, O 2017, Accelerating read atomic multi-partition transaction with remote direct memory access. in 2017 IEEE International Conference on Big Data and Smart Computing, BigComp 2017., 7881705, Institute of Electrical and Electronics Engineers Inc., pp. 239-246, 2017 IEEE International Conference on Big Data and Smart Computing, BigComp 2017, Jeju Island, Korea, Republic of, 17/2/13. https://doi.org/10.1109/BIGCOMP.2017.7881705
Murata N, Kawashima H, Tatebe O. Accelerating read atomic multi-partition transaction with remote direct memory access. In 2017 IEEE International Conference on Big Data and Smart Computing, BigComp 2017. Institute of Electrical and Electronics Engineers Inc. 2017. p. 239-246. 7881705 https://doi.org/10.1109/BIGCOMP.2017.7881705
Murata, Naofumi ; Kawashima, Hideyuki ; Tatebe, Osamu. / Accelerating read atomic multi-partition transaction with remote direct memory access. 2017 IEEE International Conference on Big Data and Smart Computing, BigComp 2017. Institute of Electrical and Electronics Engineers Inc., 2017. pp. 239-246
@inproceedings{006800a98db3424f91631fb4897355ce,
title = "Accelerating read atomic multi-partition transaction with remote direct memory access",
abstract = "Many applications these days require data processing that is both efficient and reliable. Distributed databases are one way to meet these requirements, but must be updated using distributed transactions. To manage foreign key constraints, secondary indices, and materialized views in distributed environments, read atomic multi-partition (RAMP) transactions demonstrate high efficiency. RAMP transactions achieved high performance distributed transaction processing by relaxing the isolation level. However, the use of fast interconnect to accelerate performance has not yet been considered. This paper proposes the acceleration of RAMP transactions by exploiting remote direct memory access (RDMA) on the InfiniBand interconnect. We first present GET+ and PUT+ operations that accelerate the RAMP transaction by exploiting RDMA write operations. We then present the GET∗ operation, which further accelerates GET+ operations exploiting RDMA read operations. To evaluate the proposed methods, we implemented a prototype distributed in-memory key-value store in C/C++. The results of the experiments show that compared with RAMP transactions on IP over InfiniBand, GET∗ and PUT+ achieve a 2.67× performance improvement on the Yahoo! Cloud Serving Benchmark. All of our code is publicly available.",
keywords = "Distributed Transaction, RAMP Transaction, RDMA",
author = "Naofumi Murata and Hideyuki Kawashima and Osamu Tatebe",
year = "2017",
month = "3",
day = "17",
doi = "10.1109/BIGCOMP.2017.7881705",
language = "English",
pages = "239--246",
booktitle = "2017 IEEE International Conference on Big Data and Smart Computing, BigComp 2017",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

TY - GEN

T1 - Accelerating read atomic multi-partition transaction with remote direct memory access

AU - Murata, Naofumi

AU - Kawashima, Hideyuki

AU - Tatebe, Osamu

PY - 2017/3/17

Y1 - 2017/3/17

N2 - Many applications these days require data processing that is both efficient and reliable. Distributed databases are one way to meet these requirements, but must be updated using distributed transactions. To manage foreign key constraints, secondary indices, and materialized views in distributed environments, read atomic multi-partition (RAMP) transactions demonstrate high efficiency. RAMP transactions achieved high performance distributed transaction processing by relaxing the isolation level. However, the use of fast interconnect to accelerate performance has not yet been considered. This paper proposes the acceleration of RAMP transactions by exploiting remote direct memory access (RDMA) on the InfiniBand interconnect. We first present GET+ and PUT+ operations that accelerate the RAMP transaction by exploiting RDMA write operations. We then present the GET∗ operation, which further accelerates GET+ operations exploiting RDMA read operations. To evaluate the proposed methods, we implemented a prototype distributed in-memory key-value store in C/C++. The results of the experiments show that compared with RAMP transactions on IP over InfiniBand, GET∗ and PUT+ achieve a 2.67× performance improvement on the Yahoo! Cloud Serving Benchmark. All of our code is publicly available.

AB - Many applications these days require data processing that is both efficient and reliable. Distributed databases are one way to meet these requirements, but must be updated using distributed transactions. To manage foreign key constraints, secondary indices, and materialized views in distributed environments, read atomic multi-partition (RAMP) transactions demonstrate high efficiency. RAMP transactions achieved high performance distributed transaction processing by relaxing the isolation level. However, the use of fast interconnect to accelerate performance has not yet been considered. This paper proposes the acceleration of RAMP transactions by exploiting remote direct memory access (RDMA) on the InfiniBand interconnect. We first present GET+ and PUT+ operations that accelerate the RAMP transaction by exploiting RDMA write operations. We then present the GET∗ operation, which further accelerates GET+ operations exploiting RDMA read operations. To evaluate the proposed methods, we implemented a prototype distributed in-memory key-value store in C/C++. The results of the experiments show that compared with RAMP transactions on IP over InfiniBand, GET∗ and PUT+ achieve a 2.67× performance improvement on the Yahoo! Cloud Serving Benchmark. All of our code is publicly available.

KW - Distributed Transaction

KW - RAMP Transaction

KW - RDMA

UR - http://www.scopus.com/inward/record.url?scp=85017642268&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85017642268&partnerID=8YFLogxK

U2 - 10.1109/BIGCOMP.2017.7881705

DO - 10.1109/BIGCOMP.2017.7881705

M3 - Conference contribution

AN - SCOPUS:85017642268

SP - 239

EP - 246

BT - 2017 IEEE International Conference on Big Data and Smart Computing, BigComp 2017

PB - Institute of Electrical and Electronics Engineers Inc.

ER -