SC13 Denver, CO

The International Conference for High Performance Computing, Networking, Storage and Analysis

Prov-Vis: Large-Scale Scientific Data Visualization using Provenance.

Authors: Felipe Horta (COPPE/UFRJ), Jonas Dias (COPPE/UFRJ), Renato Elias (COPPE/UFRJ), Daniel Oliveira (Fluminense Federal University), Alvaro Coutinho (COPPE/UFRJ), Marta Mattoso (COPPE/UFRJ)

Abstract: Large-scale experiments on computational engineering and science rely on compute-intensive tasks chained through a dataflow. These experiments may be modeled as scientific workflows, to ease the experiment management and take advantage of provenance data. Monitoring workflow execution is the task of verifying the status of the execution at specific points to discover if anything odd has happened. Traversing provenance data at runtime can support this monitoring, so that users can just stop or re-execute some tasks. However, most of the workflow systems execute workflows in an “offline” way. Prov-Vis is a scientific data visualization tool for large-scale workflows that uses runtime provenance queries to organize and aggregate data helping to follow the steps of the workflow and the produced partial results. A parallel numerical simulation workflow was executed on a supercomputer while Prov-Vis displayed, on a tiled-wall, visualizations of simulation steps chosen based on runtime provenance queries.

Poster: pdf
Two-page extended abstract: pdf

Poster Index