Hardware Reference
In-Depth Information
Chapter 17
ADIOS
Norbert Podhorszki, Scott Klasky, Qing Liu, and Yuan Tian
Oak Ridge National Laboratory
Manish Parashar
Rutgers University
Karsten Schwan and Matthew Wolf
Georgia Institute of Technology
Sriram Lakshminarasimhan
North Carolina State University
17.1
Motivation :::::::::::::::::::::::::::::::::::::::::::::::::::::::: 203
17.2
Design and Architecture ::::::::::::::::::::::::::::::::::::::::: 204
17.3
Deployment, Usage, and Applications ::::::::::::::::::::::::::: 205
17.3.1
Checkpoint/Restart ::::::::::::::::::::::::::::::::::::: 205
17.3.2
Analysis :::::::::::::::::::::::::::::::::::::::::::::::::: 206
17.3.3
Code Coupling ::::::::::::::::::::::::::::::::::::::::::: 207
17.3.4
Visualization ::::::::::::::::::::::::::::::::::::::::::::: 208
17.3.5
Data Reduction :::::::::::::::::::::::::::::::::::::::::: 210
17.3.6
Deployment :::::::::::::::::::::::::::::::::::::::::::::: 210
17.4
Conclusion :::::::::::::::::::::::::::::::::::::::::::::::::::::::: 211
Bibliography :::::::::::::::::::::::::::::::::::::::::::::::::::::: 211
17.1 Motivation
There are two conventional strategies for writing data from applications
that are run on supercomputers: a file-per-process, or a single-shared-file for
all processes. However, neither of these processes scale well for large-scale ap-
plications. At the application level, it is very hard to implement a flexible
I/O strategy that works well on all supercomputers. Application developers
cannot be tasked with waiting and maintaining custom I/O middleware. In
designing ADIOS, the primary goal was to create an I/O framework that en-
ables scientists to create their data-driven scientific processes, and write their
203
