Online edition of India's National Newspaper
Saturday, Apr 10, 2004

About Us
Contact Us
News: Front Page | National | Tamil Nadu | Andhra Pradesh | Karnataka | Kerala | New Delhi | Other States | International | Opinion | Business | Sport | Miscellaneous |
Classifieds | Employment | Obituary |

Business Printer Friendly Page   Send this Article to a Friend

Students' supercomputer has lessons for India

By Anand Parthasarathy

Computing: then and now

BANGALORE, APRIL 9. It failed in its stated objective — but what a magnificent failure it was! The germ of an idea mooted by San Francisco University's computer science graduate class, `CD 686', supported by its teacher, Pat Miller, became an exciting experiment in collaborative computing — and almost entered the ranks of the world's five hundred fastest super computers, last week.

The idea that one can attain supercomputer-like number crunching power by linking hundreds of ordinary desktop and laptop PCs — the type that you and I use — was proved feasible. In response to a Web invitation, nearly 700 lay PC owners turned up last weekend at the University's gymnasium, where students had readied kilometers of fibre optic cable to join everyone's machines then handed out a CD each, so that all could boot the same software and attack a common numerical problem. They called it the `Flashmob' Super Computer — the term comes from the name for spontaneous gatherings organized over Internet.

But in the end, one of the nodes failed half way through — and of the 669 machines that were finally in harness, only 256 were crunching way at a peak 180 giga flops — that's 180 million mathematical operations per second. To make it to the world's `Top 500' they should have reached 500 giga flops. Better luck next time, guys — but the idea is something the world needs — and India should be emulating for entirely sensible reasons.

(Details can be found at the project website: www.flashmo

The lesson for us is that lack of costly and cutting-edge resources need no longer be a restraint if awesome computational challenges have to be addressed.

The unused power under the hood of the average `janatha' PC, if multiplied by a sufficiently large number — like 85,000, a tiny fraction of the 8.5 million estimated PCs in India — can make for mind boggling computing power in excess of any thing that the world's fastest super computer can do. And that's the Japan-made Earth Simulator, currently clocking 35.86 teraflops (one tera flop is a trillion floating point operations).

Earlier attempts at peer-to-peer collaborative computing have attempted to enlist millions of home PCs in a quest to find a cure for cancer — but the end is nowhere in sight, possibly because it is not easy to motivate a global audience. As the San Francisco kids have shown us, things are easier when the target is smaller and the aim more modest.

Indian scenario

The Centre for Development of Advanced Computing (C-DAC) is the nodal agency for India Net: the idea is to built half a dozen supercomputers like the Param Padma (a near-1- teraflop machine that entered the Top 500 in 2003) and link them across India to tackle large problems like short range weather modelling/forecasting.

Most of these machines are yet to be paid for or built.

The Indian Institute of Information Technology and Management-Kerala (IITM-K) is the nodal agency for an ambitious Education Grid linking all institutions of higher learning in that state. Meanwhile substantial computational muscle exists — in some well equipped IITs, in the Supercomputing Education and Research Centre SERC and the CDAC's Param Padma centre, both in Bangalore, in strategic institutions like the Bhabha Atomic Research Centre (BARC), the Defence R&D's Anurag group and the National Aerospace Laboratory. The last three ran their own supercomputing initiatives — competing rather than collaborating, for over a decade — with the taxpayer forking out for these, literally — `parallel processing' attempts.

Once the dust of the Elections settles and places a new government in office, it will be high time, to relook at the nation's collective computing muscle — both `big iron' institutional machines, as well as janatha PCs, and to motivate their collaboration in quest of nationally relevant goals: a better cure for TB, an AIDS vaccine, an antidote against biological weapons. There are no lack of worthy challenges. The Indian pharma industry has already proved that it is well up to the task of drug discovery. Lack of massive computational facilities has been one perennial excuse. Last week a bunch of kids showed us that it need not be.

Printer friendly page  
Send this article to Friends by E-Mail


News: Front Page | National | Tamil Nadu | Andhra Pradesh | Karnataka | Kerala | New Delhi | Other States | International | Opinion | Business | Sport | Miscellaneous |
Classifieds | Employment | Obituary | Updates: Breaking News |

News Update

The Hindu Group: Home | About Us | Copyright | Archives | Contacts | Subscription
Group Sites: The Hindu | Business Line | The Sportstar | Frontline | The Hindu eBooks | Home |

Copyright 2004, The Hindu. Republication or redissemination of the contents of this screen are expressly prohibited without the written consent of The Hindu