Striping Layout Aware Data Aggregation for High Performance I/O on a Lustre File System

被引：2

作者：

Tsujita, Yuichi ^{[1
]}

Hori, Atsushi ^{[1
]}

Ishikawa, Yutaka ^{[1
]}

机构：

[1] RIKEN AICS, Kobe, Hyogo 6500047, Japan

来源：

HIGH PERFORMANCE COMPUTING, ISC HIGH PERFORMANCE 2015 | 2015年 / 9137卷

关键词：

MPI-IO; ROMIO; Two-phase I/O; Aggregator; Lustre;

D O I：

10.1007/978-3-319-20119-1_21

中图分类号：

TP [自动化技术、计算机技术];

学科分类号：

0812 ;

摘要：

An MPI-IO library, ROMIO, improves I/O performance for noncontiguous accesses by using its two-phase I/O optimization. When we have multiple MPI processes on each node which has multicore processors, a data aggregation assignment mismatch leads to performance degradation due to network contention. In this paper, we propose an alternative aggregation scheme that manages the striping layout of a Lustre file system to minimize network contention. The optimization has achieved up to about 30% performance improvements on our 4-node PC cluster system connected via InfiniBand FDR links in performance evaluation by an HPIO benchmark program.

引用

页码：282 / 290

页数：9

共 7 条

[1]

[Anonymous], 2008, Technical report

[2]

Ching A., 2006, P IEEE INT PAR DISTR, P49

[3]

Jialin Liu, 2013, 2013 IEEE International Conference on Big Data, P103, DOI 10.1109/BigData.2013.6691560

[4]

Kwangho Cha, 2011, Proceedings of the 2011 International Conference on Parallel Processing Workshops (ICPPW 2011), P380, DOI 10.1109/ICPPW.2011.23

[5] An abstract-device interface for implementing portable parallel-I/O interfaces [J].

Thakur, R ;

Gropp, W ;

Lusk, E .

FRONTIERS '96 - THE SIXTH SYMPOSIUM ON FRONTIERS OF MASSIVELY PARALLEL COMPUTING, PROCEEDINGS, 1996, :180-187

[6]

Thakur R., 1999, P 6 WORKSHOP IO PARA, P23

[7]

Yong Chen, 2011, Proceedings of the 25th IEEE International Parallel & Distributed Processing Symposium (IPDPS 2011), P794, DOI 10.1109/IPDPS.2011.79

← 1 →