Visual analysis of I/O system behavior for high-end computing

Abstract

As supercomputers grow ever larger, so too do application run times and data requirements. The operational patterns of modern parallel I/O systems are far too complex to allow for a direct analysis of their trace logs. Several visualization methods have therefore been developed to address this issue. Traditional, direct visualizations of parallel systems, such as Gantt charts, can be applied to parallel file systems, but do they not capture domain specific properties nor scale up to modern systems. We propose a portable I/O tracing system and visualization methods to analyze the traces we have obtained. We demonstrate the effectiveness of this system on existing parallel storage systems.