Is Kapil Sibal the next Steve Jobs?
Mr. Kapil Sibal has already produced a high flying concept called Aakash, about which, HT reports, Mr. Narendra Modi said: “The Aakash tablet promised by union minister Kapil Sibal is yet to arrive on earth.” Nevertheless, the redoubtable Mr. Sibal has already set the net ablaze with his next venture.
Telecom and IT minister Kapil Sibal is understood to have written to Prime Minister Manmohan Singh sharing the roadmap to develop “petaflop and exaflop range of supercomputers” at an estimated cost of Rs 4,700 crore over 5 years.
“In his (Sibal) letter, he has said that C-DAC has developed a proposal with a roadmap to develop a petaflop and exaflop range of supercomputers in the country with an outlay of Rs 4,700 crore,” a government official said.
A petaflop is a measure of a computer s processing speed and can be expressed as a thousand trillion floating point operations per second. Exaflop is one quintillion computer operations per second. Simply put, one exaflop is thousand times faster than one petaflop.
The fastest supercomputer in the world, Sequoia, has registered a top computing speed of 16.32 petaflops, which is equivalent of computing of power from over 7.8 lakhs high-end laptops put together.
India’s top supercomputer at present ranks 58th globally in terms of computing speed.
CSIR’s Bangalore facility has India’s fastest installed supercomputer: a platform put together by HP using off-the-shelf components. This is 1.9% of the world’s largest installation. C-DAC has India’s fifth fastest installed supercomputer, and it is only 12.9% as fast as CSIR’s in performance, so 0.2% of the world’s biggest. C-DAC is not a startup, and in its 20 years of existence, has never come close to being in the world’s top 100.
In any case, the matter of concern to us citizens is the budget. How is it possible to rebuild a company without a track record into one that makes supercomputers about 30 times faster than the world’s fastest? How is it possible to do this with a budget of Rs. 4700 crores? The answer is simple arithmetic: for this money one could buy and install a single computer system which would run at the projected speed of 1 exaflop. Presumably C-DAC plans to install a single machine which is the world’s fastest for a while. That is all. There is no question of building Indian capacity to build supercomputers.
However, there could be good science that comes out of having installed computing capacity of this kind. After all, the US government, the Japanese, the Chinese, and also India, put money into buying computers for exactly this reason. Computers bought from producers Japan and the US cannot be used for India’s defence research under the laws which apply to exports of computers from the producer countries, so Mr. Sibal’s efforts are not for the use of the defence or the atomic energy establishment. None of us is privy to India’s defense research, and how computing facilities are obtained for it, so one may as well be practical and turn our attention to industrial research, or academia.
Now India’s third largest computational facility, Computational Research Laboratory, Ltd, a wholly owned subsidiary of Tata Sons, has been trying to sell computer time to interested industries. The growth curve of their installation is interesting: they burst on the scene in late 2010 with an installed capacity of 172.60 Tflops and have added zero Tflops since. If their business model were booming, they would certainly have added capacity. One reason for this zero-growth curve could well be the current global slump. Another plausible reason is that large supercomputer users, such as Google, Amazon, and D. E. Shaw have their own captive facilities, and smaller supercomputer users have still not arisen in the Indian industrial ecology. This may change as pharma companies get into drug design in India.
India’s science financing has recognized the need for supercomputing, and in the last decade there has been a slow growth of funding to this field. The last five years, especially has been a time when large users have become larger, and many smaller supercomputing users have emerged. This is a superb growth model: small users train themselves into being larger users, and keep hiving off other small user groups. Evidence of this is the slow rise in academic supercomputing facilities seen in the Top Supercomputers, India list, and a general growth in computing budgets across academic departments throughout the country. The academic strength of this emerging computational science community (not computer science, which is a different beast) needs to be bolstered not only by direct funding into computing resources, but also by a sustained institutional support to grow and nurture manpower. With a larger academic training and manpower scheme, it is inevitable that this will fuel growth in the growing industrial research sector in the future. The government has to look to academic training, especially in areas like computational science where India has basic strength, as a medium-term investment into industrial growth.
Unfortunately, C-DAC has never been part of this solution. It has not bothered to open its facilities to science applications, or perhaps not been professional enough to guarantee that the facility will run 24 hours a day, 365 days a year, with enough attention paid to connectivity and storage of large data sets, and with user consultants who are available around the clock. Given its track record, C-DAC probably has not yet come to address the question of how to ensure a continuous supply of the 400 Megawatts of power required to run its exaflops system. So by taking this organization as his flagship for the project, Mr. Sibal has ensure that his efforts will turn out to be just a media circus, of no lasting value to India.