High-Performance Parallel Simulation of Fluid Dynamics on Cluster Architectures Using MPI

International Journal of Computer Science and Technology | Vol. 7, No. 5, May 2005 | pp. 33–40

DOI: 10.46882/2005/IJCST/070508

Research Article

Title: High-Performance Parallel Simulation of Fluid Dynamics on Cluster Architectures Using MPI

Names of Authors: Guillaume Mercier¹, Aparna Rao²

Authors’ Affiliations: ¹Département d'Informatique, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland; ²Department of Computer Science and Automation, Indian Institute of Science, Bangalore, India

Abstract:
Simulating complex fluid dynamics variables requires significant computational resources, often necessitating high-performance parallel computing solutions. This paper implements a parallelized Navier-Stokes equation solver across distributed-memory cluster setups using the Message Passing Interface (MPI) framework. The computational space is divided using a 3D block-decomposition algorithm designed to minimize border communication overhead between processing ranks. Evaluation on a 64-node Linux cluster system achieved a parallel execution speedup efficiency metric of E = 87.5% when scaling up to maximum core allocations. The processing latency added by inter-node network communication was kept exceptionally low at less than 6.2% of total operational runtimes. This was achieved by using non-blocking MPI communication primitives (MPI_Isend and MPI_Irecv) to overlap computational steps with network data transfers. The framework can compute complex mesh structures exceeding 10⁸ elements, providing a scalable backbone for industrial aerodynamic design testing.

Keywords: Parallel Computing, Fluid Dynamics, MPI, Domain Decomposition, Scalability, Computational Grids

Manuscript Timeline: Received February 10, 2005; Revised March 24, 2005; Accepted April 15, 2005; Published May 26, 2005

Citation: Mercier, G., & Rao, A. (2005). High-Performance Parallel Simulation of Fluid Dynamics on Cluster Architectures Using MPI. International Journal of Computer Science and Technology, 7(5), 33–40.

Subscribe to read the full article: https://internationalscholarsjournals.org/subscribe-to-read