The short answer is: India is not yet ready for serious HPC (high performance computing) business. TCS has scooped up a great start-up which will allow it to offer more differentiated services to its global clients.
As most media outlets have reported today, TCS is acquiring Pune-based Computational Research Laboratories (CRL) for Rs 188 crore.
Here’s a company that was set up in 2007, never mind with a storied beginning, to be the first corporate-funded HPC facility in a world that is dominated by government-funded supercomputing machines. The synergy between some of the Tata group companies was obvious: they would aid and use CRL expertise. But CRL, as officials told me then, was supposed to prove that supercomputing can be a viable business by itself. That, as this acquisition shows, didn’t quite pan out as the group had planned: Tata Sons, whose wholly owned subsidiary CRL was, had invested $30 million then, and now TCS is paying $34 million to acquire it!
I got a whiff of the changes at CRL a few weeks ago when I learnt that its chief executive Vipin Choudhary is going back to academics in the US, where he came from to lead CRL. With both academic and entrepreneurial experience, Chaudhary was expected to drive the business.
When I asked Chaudhary in late July if he was going back to theUS, he said he had always traveled between India and US on business and his family continues to be there. Today he says he’ll continue to work for TCS, and “this is a big win for CRL and HPC”.
With most of its customers overseas, CRL was finding it tough to break into the Indian market which is not only small but also ridden with turf war among institutions that have an unhealthy habit of putting together big machines without thinking through its effective use.
When CRL started, it hoped to bundle its brute force capacity with novel applications to serve various Indian government programs – space, climate research, weather forecasting, life sciences, media and entertainment industry, etc. But, leader as CRL is in HPC cloud and solutions space, according to Chaudhary, there are very few government projects of this nature.
“Perhaps most projects you have seen are hardware acquisition which comes from MNCs and the companies who supply them do not have any value add. That is not our model,” says Chaudhary.
Back in 2008, S Ramadorai, chairman of CRL and CEO of TCS had told me that the challenge for CRL was ground-up. “Unless we involve the research community, unless the industry has a real world problem, or we redefine the problems, for instance, improvement in manufacturing at the nanoscale level, penetration of HPC will take time,” said Ramadorai, who was then talking to Bharat Heavy Electricals Ltd for deploying Eka’s (name of CRL’s supercomputer) services, among other things, for better power transmission and distribution. How much of such business deals were clinched in India, is anybody’s guess.
Eka is still one of the largest systems in India after five years. It is the largest commercial system by a big margin and perhaps close to top five in the world in that category, argues Chaudhary.
Incidentally, in June this year the government announced Rs 5000 crore for ramping up supercomputing infrastructure in the country.
Ashwani Kumar, Union Minister of State for Planning, Science and Technology and Earth Sciences, said, “The Indian government is seriously looking at accelerating R&D in HPC infrastructure development and technology proliferation. For this, an allocation of Rs 5,000 crore has been made in the 12th Plan.” Several supercomputing facilities in a “three-tier architecture” which would include many small-scale HPC systems at various institutes, some medium-scale HPC systems at regional level, and a few large-scale HPC systems at the national level would be set up.
Ideally, the govt should make public what it plans to achieve, in measurable and effective outcomes, with Rs 5000 Cr investment. Supercomputing after all is pretty commoditized today. What it actually boils down to are smart applications, and how much juice one is able to extract out of the fast number crunching ability. A fresh impetus to supercomputing in India shouldn’t just become a typical “big science” project which is undertaken for stature and, in many cases, for survival of research groups.