CFIO: A Fast I/O Library for Climate Models

Author

Wencan Wang ; Xiaomeng Huang ; Haohuan Fu ; Yong Hu ; Shizhen Xu ; Guangwen Yang

Author_Institution

Center for Earth Syst. Sci., Tsinghua Univ., Beijing, China

fYear

2013

fDate

16-18 July 2013

Firstpage

911

Lastpage

918

Abstract

How to improve I/O performance is a significant challenge for modern scientific applications. In this paper, we present Climate Fast I/O (CFIO), a parallel I/O library that is specifically developed for climate models. To minimize the required code modifications for switching to the CFIO, we provide the same interface and feature as PnetCDF and adopt an I/O forwarding technique to provide automatic overlapping of I/O with computing. The experimental results show that CFIO performs better than PnetCDF in terms of decreasing the overall running time of the program. By using the CFIO, the 0.1 degree POP ocean model that runs on one thousand CPU cores can be accelerated by 7.9 times over the original version.

Keywords

climatology; geophysics computing; input-output programs; software libraries; CFIO performance; CPU core; I/O forwarding technique; POP ocean model; PnetCDF; climate fast I/O; climate model; modern scientific application; parallel I/O library; Computational modeling; Data models; Libraries; Meteorology; Oceans; Servers; Writing; Climate Models; I/O forwarding; POP; Parallel I/O;

fLanguage

English

Publisher

ieee

Conference_Titel

Trust, Security and Privacy in Computing and Communications (TrustCom), 2013 12th IEEE International Conference on

Conference_Location

Melbourne, VIC

Type

conf

DOI

10.1109/TrustCom.2013.111

Filename

6680931