Case studies of solving large-scale CFD problems by means of the GasDynamicsTool software package

Abstract

Numerical simulations of various CFD problems typically require extra-large computational domain usage with the number of modeling cells in the tens of thousands. The task of simulating such large models is one which many engineers and developers face, and is often a sticking point for many reasons. That's the motive for GDT Software Group to consider the development and optimization of parallel high performance modeling tools for large-scale tasks and multiple-processor systems as one of its first-rank priority goals [1]. The package is designed and is being continually optimized to provide the ability to simulate problems with 1-10 billions of computational cells effectively on ordinary cluster systems. The article describes attempts being made to improve software quality and performance on cluster systems with the number of CPUs up to 200 and some achievements made in production problem-solving on such systems. © 2009 Springer-Verlag Berlin Heidelberg.