The International Conference for High Performance Computing, Networking, Storage, and Analysis
The International Conference for High Performance Computing, Networking, Storage, and Analysis
Workshop:ISAV21: In Situ Infrastructures for Enabling Extreme Scale Analysis and Visualization
Authors: Joao Barbosa (University of Minho, Portugal); Paul Navratil (University of Texas at Austin); Luis Paulo University of Minho, Portugal (Universidade do); and Donald Fussell (University of Texas at Austin)
Abstract: Post-hoc high-fidelity scientific visualization (HSV) of numerical simulations (NS) requires multiple I/O check-pointing to inspect the simulation progress. The costs of these I/O is high and can grow exponentially with increasing problem sizes. In situ HSV dispenses with costly I/O, but requires additional computing resources (CR) to generate the visualization, increasing power and energy consumption. In this paper we present LOOM, a new interweaving approach supported by a task scheduling framework to allow tightly coupled in situ visualization without adding to the NS runtime. The approach exploits idle time of the NS, due to workload imbalances, to perform the visualization. Execution time and power requirements are minimized by sharing the same CR among NS and visualization. We demonstrate that LOOM reduces time to visualization by 3x compared to a traditional non-interwoven pipeline. Our results here demonstrate good potential for additional gains for large distributed-memory use cases with larger interleaving opportunities.