SC19 Proceedings

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

Poster 63: Adaptive Execution Planning in Biomedical Workflow Management Systems

Authors: Marta Jaros (Brno University of Technology), Bradley E. Treeby (University College London), Jiri Jaros (Brno University of Technology)

Abstract: Biomedical simulations require very powerful computers. Their execution is described by a workflow consisting of a number of different cooperating tasks. The manual execution of individual tasks may be tedious for expert users, but prohibiting for most inexperienced clinicians. k-Dispatch offers a ‘run and forget’ approach where the users are completely screened out from the complexity of HPC systems. k-Dispatch provides task scheduling, execution, monitoring, and fault tolerance. Since the task execution configuration strongly affects the final tasks mapping on the computational resources, the execution planning is of the highest priority. Unlike other tools, k-Dispatch considers a variable amount of computational resources per individual tasks. Since the scaling of the individual HPC codes is never perfect, k-Dispatch may find such a good mapping even an experienced user would miss. The proposed adaptive execution planning is based on collected performance data and the current cluster utilization monitoring.

Best Poster Finalist (BP): no

Poster: PDF
Poster summary: PDF

Back to Poster Archive Listing