Scalability and Performance of CTH on the Computational Plant
Ron Brightwell, Haijie Fang, Lee Ward · 2006
| In this paper, we present performance results from CTH, a family of codes developed at Sandia National Laboratories for modeling complex multi-dimensional, multi-material problems characterized by large deformations and/or strong shocks. We compare results from the 400-node Computational Plant (Cplant TM ) Linux cluster to performance obtained on traditional supercomputing platforms: Intel TeraFLOPS, IBM SP-2, and a cluster of Digital Equipment Corporation 8400 servers. We provide an overview of the Cplant TM platform and system software environment, and discuss the characteristics of the machine that inuence performance and scalability of the application. I. Introduction Parallel computing platforms composed of commodity personal computers (PC's) have become an attractive alternative to traditional supercomputing platforms. Recently, procurements for small- and medium-sized parallel computers have been bid on, and in some cases have been awarded to, commodity-based PC cluste...