Benchmarking Parallel IO Performance for Computational Fluid Dynamics Applications资料.pdf

43rd AIAA Aerospace Sciences Meeting and Exhibit, 10 - 13 Jan 2005, Reno, Nevada Benchmarking Parallel I/O Performance for Computational Fluid Dynamics Applications ∗ † Parimala D. Pakalapati Thomas Hauser Utah State University, Logan, Utah, 84322, USA I. Introduction Linux clusters can provide a viable and more cost effective alternative to conventional supercomputers for the purposes of computational fluid dynamics (CFD). In some cases, the Linux super cluster is replacing the conventional supercomputer as a large-scale, shared-use machine. In other cases, smaller clusters are providing dedicated platforms for CFD computations. One important, often overlooked, issue for large, three dimensional time-dependent simulations is the input and output performance of the CFD solver. The development of the CFD General Notation System (CGNS) (see1, 2, 3 ) has brought a standardized and robust data format to the CFD community, enabling the exchange of information between the various stages of numerical simulations. This paper presents benchmarks results of the parallel I/O performance of a proposed parallel CGNS extension.4 II. The parallel CGNS system The specific purpose of the CFD General Notation System (CGNS) project is to provide a standard for recording and recovering computer data associated with the numerical solution of the equations of fluid dynamics. The intent is to facilitate the exchange of Computational Fluid Dynamics (CFD) data between sites, between applications codes, and across computing platforms, and to stabilize the archiving of CFD data. The CGNS system consists of two