Extending CoMedia: A Service Architecture for Grid Computing in Automotive Design 2001-01-3465

Grid Computing is a relatively new concept that provides a high performance metacomputing platform consisting of multiple computers connected using high-speed networks. The metaphor computational grid is based on the electrical power grid, which provides ubiquitous electrical power. In the same way, a computational grid has the potential of providing ubiquitous computing power. Connection of many computers at appropriate speeds has become more feasible with the recent deployment of high-speed second-generation optical Internet infrastructures, such as Internet2 and CA-Net 3 in North America and DANTE/ GÉANT in Europe. These networks can interconnect computing nodes in a grid at up to gigabit speeds.

This paper is an exploration of the potential of using Grid Computing services to serve the Automotive Design industry. The researchers first report on CoMedia, a joint Canada-Germany research initiative to provide multimedia and visualization services of use to automotive and industrial designers. The research team will report on extensions of this service architecture based on the principles of Grid Computing. Initially, large clusters of linux-based systems have been utilized to provide compute-intensive services such as 3D rendering using MPI. The paper will discuss an approach to utilizing spare compute resources available on a grid by use of intelligent agent technology such as aglets, which promise to improve node utilization based on node processor speed, spare cycles available or other important parameters.

Finally, the paper will discuss the types of services that can be made available to designers, including visualization and modeling, as well as other compute-intensive methodologies such as finite element analysis and computational fluid dynamics. The economics and performance of such a system will conclude the paper.