High-Performance Cluster-Based Computing for Engineering Design

A small two-day conference arranged by SAE India (Western Section) just got over in Pune today… The conference was arranged as a part of the annual (or so) auto industry/trade exhibition called “AutoDesign”. The trade exhibition itself was held at the ICC (International Convention Center) in Pune, on the 16th and 17th of February, whereas the more specialized or technically oriented conference to go with it was held later on, on the 18th and 19th of February, at the ARAI (Automotive Research Association of India), Pune.

I was invited to deliver a talk in this conference, and I gave an introductory one on the title topic. (This is precisely the conference that I had mentioned I was preparing for, in one of my earlier posts below.)

I had thought that the title topic would be relevant for this year’s “AutoDesign.” I thought that the news of Tata Sons’ “Eka” being the 4th fastest supercomputer in the world, would have generated some excitement even among the automobile engineering student community. But, once at the conference, I found out that I was somewhat off the mark in keeping that anticipation. The audience was not so well-geared towards hearing much about mega-flops and giga-flops. My talk did not quite go flat as such… But still, there wasn’t much excitement about what a faster computer could do for them—the very subject matter itself….

There seems to be this gap or disconnect between the three kinds of people involved: the automobile styling professionals, the mechanical design engineers, and the recent advancements in computers field. This was a bit surprising to me for a city like Pune. In a way, of course, that kind of a disconnect precisely highlights the relevance and the necessity of the events like AutoDesign.

Anyway, I am now attaching the .PDF slides of my own talk here:  AjitJadhav_SAE_AutoDesign2008.PDF (219 kB). This PowerPoint presentation contains some introductory remarks as well as some preliminary but honest data about how the ordinary PC desktops would actually fare when they are used as individual nodes in an HPC cluster. That is, for real-life sized data (i.e. for larger matrices)…. Feel free to download the PDF file of the slides, and do leave a comment if you think that I am making a mistake somewhere in them.