Giga-hit or mega-flop?

Will the fastest supercomputer developed by India help the country realise its potential? Or will it flop?

Last Updated: Saturday 04 July 2015

  Standing tall: PARAM 10000 c if one goes by precedents, this surely is a success that Indian hardware engineers can boast of. The Centre for Development of Advanced Computing ( c-dac ), Pune, recently unveiled the param 10000, its latest supercomputer. param 10000 can perform 100 billion (100,000,000,000) mathematical operations per second, making it Asia's second fastest computing machine.

The announcement of the development was well timed, coming as it did just when the us , the largest exporter of supercomputers, announced further control over export of fast supercomputers to third world countries like India. In fact, 10 years ago the us had refused to sell the Cray ymp supercomputer to India on the excuse that it could be used for military purposes. Since c-dac came up with param 10000 six months ahead of schedule, it pre-empted moves by the us government and industry to retain control over the market for supercomputers.

But, as far as uniqueness of architecture goes, this is just a development over the previous param versions developed by c-dac . Moreover, in terms of supercomputing capabilities, computers in the us can outpace the machine by 10 times. So is brouhaha over the development justified?

All computers ultimately work on mathematics in binary notation (0s and 1s). Using this notation, a computer can work with true-false conditions and sort out logical problems. But mathematical problems requiring endurance -- say, calculating the value of pi (p) to the millionth position to the right of the decimal -- are not fed into a computer using binary notation. Rather, a variety called the floating point system is used, where calculations are done with numbers where the numerals to the right of the decimal are not fixed. The kinds of problems that supercomputers tackle are, therefore, measured in terms of floating point operations per second (or flops). The fastest machines have a processing capability measured in tera flops (trillion operations per second). param 10000 has a capability of 100 giga (billion) flops, but can be upgraded to tera-flops. The fastest computer being developed in India before the arrival of param 10000 was a 30 giga-flop machine to be unveiled by the Defence Research and Development Organisation ( drdo ) by August 1998.

The c-dac machine uses the principle of parallel processing -- a type of multiprocessing where the work is not just divided into chunks between different processors, but that parts of the same mathematical problem are solved simultaneously by a number of processors. This is much like solving arithmetic problems using the unitary system. The Cray and ibm supercomputers, on the other hand, use sequential processing. Both types of processing have their advantages, but the market for computers with parallel processors is small. Demand has been further dampened by us competitors who have reportedly been 'dumping' computers at below-market prices. But this in itself is a measure of India's success, whose entry into the market with supercomputers at just Rs 50 crore, has shaken the American majors.

Unlike supercomputers developed by large organisations -- Silicon Graphics-Cray, ibm and Hewlett-Packard -- which spend billions of dollars developing housekeeping software for their systems, param 10000 needs to be customised each time it is installed. In other words, though the hardware is excellent for such a low price and is indigenously developed, to use the computer an organisation needs to spend on software development.

For mega corporations this is not much of a problem. In fact, the c-dac machine could go a long way in meeting the requirements of Indian giants like the Indian Space Research Organisation ( isro ), drdo and the Oil and Natural Gas Commission ( ongc ), apart from the Indian Meteorology Department. On the flip side, tailor-made software has an advantage -- it offers more security.

The possibilities of use of param 10000 in military development -- simulating nuclear explosions, miniaturisation of nuclear devices, reduction of the missile development cycle -- have been much talked. But its uses in civil life, especially weather forecasting, satellite development, oil and natural gas exploration, and molecular modelling in genetics and drug development are immense. The one problem that clouds the prospects is loss of software engineers by the c-dac in particular, and the industry in general. Most of these have left for greener pastures abroad, mainly to the us . But many brilliant people have stayed back. Chances are that the full potential of param 10000 and other such developments will be realised, sooner than later.

Subscribe to Weekly Newsletter :

Comments are moderated and will be published only after the site moderator’s approval. Please use a genuine email ID and provide your name. Selected comments may also be used in the ‘Letters’ section of the Down To Earth print edition.