Hardware Reference
In-Depth Information
28.3 Conclusion
Our application of Iota to I/O kernels demonstrate that complete tracing
of POSIX I/O calls is possible for petascale MPI applications. In addition, by
querying Lustre metadata during tracing, we can reconstruct the mapping of
I/O operations to individual OSTs when post-processing the traces.
Bibliography
[1] Surendra Byna, Jerry Chou, Oliver Rubel, Prabhat, Homa Karimabadi,
William S. Daughton, Vadim Roytershteyn, E. Wes Bethel, Mark Howison,
Ke-Jou Hsu, Kuan-Wu Lin, Arie Shoshani, Andrew Uselton, and Kesh-
eng Wu. Parallel I/O, Analysis, and Visualization of a Trillion-Particle
Simulation. In International Conference for High Performance Comput-
ing, Networking, Storage and Analysis (SC12), Salt Lake City, UT, USA,
November 2012.
[2] Philip Carns, Kevin Harms, William Allcock, Charles Bacon, Samuel Lang,
Robert Latham, and Robert Ross. Understanding and Improving Com-
putational Science Storage Access through Continuous Characterization.
ACM Transactions on Storage, 7(3), October 2011.
[3] Philip Carns, Robert Latham, Robert Ross, Kamil Iskra, Samuel Lang,
and Katherine Riley. 24/7 Characterization of Petascale I/O Workloads.
In Workshop on Interfaces and Abstractions for Scientific Data Storage
(IASDS '09), New Orleans, LA, USA, September 2009.
[4] Mark Howison, Quincey Koziol, David Knaak, John Mainzer, and John
Shalf. Tuning HDF5 for Lustre File Systems. In Workshop on Inter-
faces and Abstractions for Scientic Data Storage (IASDS '10), Heraklion,
Crete, Greece, September 2010.
[5] Andrew Uselton, Mark Howison, N.J. Wright, David Skinner, Noel Keen,
John Shalf, K.L. Karavanic, and Leonid Oliker. Parallel I/O Performance:
From Events to Ensembles. In IEEE International Symposium on Parallel
& Distributed Processing (IPDPS '10), Atlanta, GA, USA, April 2010.
