Saturday, October 16, 2004

What a Prerna
Minna Zutshi on the three Jalandhar compkids who have developed a supercomputer

Sandeep Kath (in spectacles), Abhinav Dhall (centre) and Vibhore Kumar Jain
Sandeep Kath (in spectacles), Abhinav Dhall (centre) and Vibhore Kumar Jain

If you think success is heady and it trills its way to smugness, take some time off to meet barely-out-of-teens B Tech Computer Science Part III students Abhinav Dhall and Vibhore Kumar Jain and their twenty-something teacher-mentor Sandeep Kath. These youngsters, all from DAV Institute of Engineering and Technology, Jalandhar, may have developed a supercomputer "Prerna" but their feet are firmly on the ground. Yet they are not resting on their laurels. With less than three months into "supercomputer-creation", they are already raring to go many GigaFlops.

"Yeh dil mange more. Our next step would be to beat the Param, the fastest supercomputer in our country," sallies Abhinav in a voice suffused with conviction. His making it sound so simple and effortless has you rooting for more bytes.


* Prerna is a 25-node Linux cluster
* Developed at Prisma (software development cell), DAV Institute of Engineering and Technology, Jalandhar
* Connected to 23 nodes
* Speed is 48 GFlops
* Currently used for parallel processing and SETI (Search for Extra-Terrestrial Intelligence)@Home
* Can be used for MATLAB and fibre-optics simulation

His friend, Vibhore, chips in, "Prerna was one month’s hard slog. It was10x30x3 hours of focused work - all three of us put in 10 hours of tough grind daily for a month. We brooked no distractions during this time." His favourite authors Mark Twain and Robert Frost were abandoned for a month, as his mind was occupied with "clusters and nodes" those days. "A cluster is a NOW (Network of Workstations) performing a complex task. It is a group of computers solving problem collectively, using 'divide and conquer methodology'. Clusters seem to be a promising solution to the quest for speed," explains Kath, adding, "If we utilise IT in agricultural sector, our production is bound to increase by 30 per cent."

For these youngsters, computers did not "happen" out of the blue. They had grown up sampling hot, snazzy bytes of computers
(real and on-the-verge-of-reality). "When I was a kid, I dreamt of DNA computers, complete with all their imagined features. Comps fascinated me no end," Vibhore says.

"It's a common misconception that supercomputers have something very geeky about them. Also, people think supercomputers are prohibitively expensive. The basic funda is about networking. We worked with the hardware already available at our college. We joined the systems - made one the CEO and the rest the workers," elaborates Abhinav.

Their talk is punctuated with GFlops, TFlops, fireballs and yet more comp lingo. But it's not that their minds are always clustered with bytes. They love to have dollops of fun (movies, music et al) too.

Last week, they met scientist Y.S. Rajan, who exhorted them to cross the theoretical output of 1 TFlops of the Param. "Right now, the benchmark should be to go beyond this speed," they were told. All three of them consider IT research as their ultimate objective.

Interestingly, when Abhinav broke the news of developing a supercomputer to his parents, his father's spontaneous reaction was: "What exactly is it? This means you'll stop studying."