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

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 x on some samples for both of the target systems.

I. Introduction

In HPC there is a need for tools to predict I/O performance accurately and interpret factors that influence performance. For supercomputer facilities, more predictable I/O performance enables more precise core-time allocations and more efficient system utilization. For domain scientists with limited budgets for compute cycles charged on application runtimes [1], accurate prediction offers the potential to reduce I/O-induced idle time in their runs (§II-A1, §IV-D).