Optimizing parallel CFD simulations for modeling a trimaran through OpenFoam

Abstract

Computing of fluid dynamics is gaining popularity over the last years and requirements such as accuracy and speed are getting bigger. Software products that can handle much more complex tasks for the industry, than the past are already available on the market. One of the problems for the industry becomes how a given software for computational fluid dynamics (CFD) simulation could be parallelized and optimized to achieve better performance when solving complex engineering problems. To predict the hydrodynamics forces and moments requires hours, even a days of work of cluster platforms with large computing power. The biggest cluster platforms contain enormous number of cores and large amounts of memory, and have the capability of solving complex problems for a few minutes. In this paper we study a parallel CFD simulations for modeling a trimaran through OpenFoam CFD toolbox. Results for scalability of using MPI parallel model with/without hyperthreading (HT) are presented. The performance of the CFD simulations is done on a supercomputer platform Avitohol [3]. The scalability criterion is introduced. The parallel CFD simulation testsfor modeling a trimaran were done to demonstrate the parallel performance of the OpenFoam software and to define efficient usage of Avitohol’s resources based on the introduced scalability criterion. Guidelines for future work are also presented.