An I/O Mini-App Dedicated to In Situ Visualization

Abstract

In situ visualization promises to offer one solution to the problem of stagnant I/O bandwidths relative to computing capacity. Yet, it has drawbacks, including a lack of explorable results. This can be addressed by producing derived or extracted results instead of just images. However, the I/O for these results is not guaranteed to be sufficiently scalable if not designed properly. We present an I/O mini app that measures derived quantity I/O performance. It produces isosurfaces with a data generation function allowing fine control of the volume, load imbalance, and other aspects of the isosurface geometry output data. We also provide ongoing results of scalability benchmarks with various output methodologies.