Numerically Intensive Simulations on an Integrated Compute Infrastructure

Scientific simulations are often computation intensive and time consuming and can highly profit from choosing a suitable specialized computing resource. However, choosing the right computing resource is not a trivial task, especially in modern computing centers, that offer access to a heterogeneous infrastructure including cloud services, high performance computing clusters and clusters with specialized accelerators (e.g. GPU cards). Therefore, in this project, a concept and implementation of a transparent integration mechanism for heterogeneous computing resources is developed. The goal is to semi-automatically deploy and execute simulation applications on the most suitable resource type.