Abstract
Keywords: CAE, IT, Grid, Computer, Crashworthiness
Abstract
In order to improve automotive safety, complex types of investigations to vehicle crashworthiness must be checked in design stages, and numbers of crash simulations in the design stage have increased more and more. Computer resources for crash simulations are requested to be in better price/performance.
When computer resources for the CAE usage in automotive development are necessary to be procured, CAD workstations in idle periods, midnight and weekend, had been a candidate for a long time ago. Recent popular term grid computing is expected that the networked resources show a lot of application possibilities in future. Author had continued to realize such trials for many years. Several inhouse codes such as a CFD code were tried and tuned in the level of source codes in order to show good performance on the networked servers. But the ability of networked resources could not reach the ability of dedicated vector computers in 1996.
Nowadays, however, workstations and networks have improved performance. Based on current improved networked environments, we develop a grid system on the cluster of CAD workstations for crash simulation. It works as CAE server simultaneously driven in idle periods of CAD usage, and its ability is comparable to that of supercomputer. 12 workstations with 100 M bit/s network can show a performance equivalent of vector computers.
A system to control batch jobs has been developed, without introducing a code of general purpose. Networked resources are exposed to various artificial environments even in idle periods. Our inhouse system is able to adopt practical irregular environments.
The cluster system is put in practical use. There have occurred subjects in actual employment in the EWS cluster for crash simulation. Some subjects are conquerable, but some subjects can not be solved under current grid technology. Details are explained and discussed in the paper.