Interpreting Write Performance of Supercomputer I/O Systems with Regression Models

Conference · Sat May 01 04:00:00 EDT 2021

DOI:https://doi.org/10.1109/IPDPS49936.2021.00064· OSTI ID:1809965

Xie, Bing ^[1]; Tan, Zilong ^[2]; Carns, Philip ^[3]; Chase, Jeffrey ^[4]; Harms, Kevin ^[3]; Lofstead, Gerald ^[5]; Oral, Sarp ^[1]; Vazhkudai, Sudharshan S. ^[6]; Wang, Feiyi ^[1]

---

1. ORNL
2. Carnegie Mellon University (CMU)
3. Argonne National Laboratory (ANL)
4. Duke University
5. Sandia National Laboratories (SNL)
6. Micron Technology Inc

This work seeks to advance the state of the art in HPC I/O performance analysis and interpretation. In particular, we demonstrate effective techniques to: (1) model output performance in the presence of I/O interference from production loads; (2) build features from write patterns and key parameters of the system architecture and configurations; (3) employ suitable machine learning algorithms to improve model accuracy. We train models with five popular regression algorithms and conduct experiments on two distinct production HPC platforms. We find that the lasso and random forest models predict output performance with high accuracy on both of the target systems. We also explore use of the models to guide adaptation in I/O middleware systems, and show potential for improvements of at least 15% from model-guided adaptation on 70% of samples, and improvements up to 10× on some samples for both of the target systems.

• Research Organization: Oak Ridge National Laboratory (ORNL), Oak Ridge, TN (United States)
• Sponsoring Organization: USDOE
• DOE Contract Number: AC05-00OR22725
• OSTI ID: 1809965
• Country of Publication: United States
• Language: English