Abstract

There is a clear desire on the part of a number of domain and computer scientists to see a convergence between the two high-level types of computing systems, software, and applications: Big Data Analytics (BDA) and High Performance Computing (HPC). In some sense scientists do not care how they obtain their computational results, as long as they are generated in a timely and robust manner. Some applications naturally cross the boundaries between HPC and BDA, for example executing a large-scale simulation on an HPC system and then performing analytics on the results (either ex-situ—on another computational platform, or in-situ—within the HPC system.
Up to now, workflow management solutions have been developed to help cross the computational system boundaries. Workflows can be composed of a variety of computational tasks: tightly coupled codes, machine learning loosely coupled applications …

Date

2016

Authors

Ewa Deelman

Journal

Big Data and Extreme-Scale Computing