Resources
Our initial DLIO paper published in CCGrid’2021 described the design and implementation of DLIO benchmark.
@article{devarajan2021dlio,
title={DLIO: A Data-Centric Benchmark for Scientific Deep Learning Applications},
author={H. Devarajan and H. Zheng and A. Kougkas and X.-H. Sun and V. Vishwanath},
booktitle={IEEE/ACM International Symposium in Cluster, Cloud, and Internet Computing (CCGrid'21)},
year={2021},
volume={},
number={81--91},
pages={},
publisher={IEEE/ACM}
}
DLIO is the key software for the MLPerf Storage benchmark: https://mlcommons.org/en/groups/research-storage/. See also the following relevant paper from MLPerf Storage working group:
@article{balmau2022mlperfstorage,
title={Characterizing I/O in Machine Learning with MLPerf Storage},
author={O. Balmau},
booktitle={SIGMOD Record DBrainstorming},
year={2022},
volume={51},
number={3},
publisher={ACM}
}