bringing back india’s supercomp

Upload: ashish-krishna

Post on 04-Jun-2018

222 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/13/2019 Bringing Back Indias Supercomp

    1/3

    Publication: The Economic Times Mumbai; Date:2011 Aug 18; Section:Technology; PageNumber: 4

    Teraflop Turf: Bringing Back Indias Supercomp

    FUTURE TECH: Once a leader and then a laggard, India is pushing hard theenvelope on high-performance computing

    HARI PULAKKAT BANGALORE

    On December 13 last year, N Balakrishnan brought together 28 of Indias most resourceful

    minds in the field of supercomputing. Balakrishnan is the associate director of the Indian

    Institute of Science (IISc) in Bangalore, an aerospace engineer, and a voracious consumerof supercomputing resources. He had been asked by a worried Planning Commission to

    brainstorm about Indias supercomputing landscape. India was one of the world leaders inthis tech not long ago, but had been left far behind by other countries within a few years.

    China had come up from a nearabsent position to the top within a decade, now rivaling the

    US and Japan, and in a good position to overtake them soon in supercomputing resourcesand knowhow. There is a strong correlation between a countrys access to supercomputing

    facilities and its achievements in science and technology, says Balakrishnan.

    Scientists who were potential users of supercomputing had taken the situation seriously. Sohad the Planning Commission when its members were told about Indias decline in this field.

    After several meetings and a conference, the Scientific Advisory Committee to the Prime

    Minister (SAC-PM) recommended a national programme on supercomputing at a cost of Rs5-6000 crore. Several leading scientists had, in turn, made similar recommendations to the

    SAC-PM. Some of them were grand visions to turn India into a supercomputing hub, butmany envisaged using the programme to develop cutting edge commercial tech as well.

    Balakrishnan himself had recommended a large project that would involve developing

    several supercomputers of varying speeds. The Planning Commission agreed in principle tofund the project, the level of which could go up to Rs 10,000 crore if necessary.

    So by the end of the year, there might start a project that would be the largest that India

    has ever undertaken outside the realm of defence, atomic energy and space. It would use

    mostly off-the-shelf chips but develop cutting-edge technologies in many other areas. Theproject would involve the private sector in a big way and thus could kick-start several newcompanies. It would network Indias scientists and engineers at a scale never achieved so

    far. Finally, when the first phase of the project is complete within four to five years, it would

    place enormous computing resources within easy reach of Indias scientific community and

    the private sector. The success of such a project would also depend on its commercialviability, says Vijay Bhatkar, founder-director of the Centre for the Development of

    Advanced Computing (C-DAC), and one of the chief architects of India's first foray intosupercomputing. So they would need to be general purpose machines capable of solving

    industrial problems as well.This was not the case earlier, but C-DAC developed its supercomputers at a time when

    space and defence were their primary users. Supercomputing has now moved to the

    mainstream, with a large number of applications in many areas of science and technology.

    Parts of modern biology research depend on supercomputing for their survival. Humangenome sequencing, for example, wouldnt be possible without the availability of

    supercomputers to analyse the results; and sequencing is rapidly becoming an integral partof clinical drug development.

    Climate researchers desperately need immensely-powerful

    machines; even weather forecasters need better stuff than they use now. So do materialscientists, aerospace engineers, mathematicians and several other researchers. As

    technology advances, says R Govindarajan, professor and chairman of the Supercomputer

    Education and Research Centre (SERC) at IISc, we ask for more and more resolution in ourresults. Each higher resolution needs the use of more powerful computers.

    Use of supercomputers is expanding rapidly in the private sector too. Analysis of data fromsocial networks sometimes needs supercomputers and would stretch the most powerful ones

    as data volumes increase in the future. Video analytics also needs high-speed computers,and so would data search, crash analysis, 3D animation, business data analytics and other

    every day business activities. The number of nodes in a crash analysis now runs intomillions, says Subram Natarajan, executive of deep computing, IBM India and South Asia.

    Page 1 of 3Teraflop Turf: Bringing Back Indias Supercomp

    1/30/2012http://epaper.timesofindia.com/APA26300/PrintArt.asp?SkinFolder=pastissues2

  • 8/13/2019 Bringing Back Indias Supercomp

    2/3

    IBM recently used supercomputers to improve the productivity of an Indian oil company. In

    fact, supercomputers are so fundamental to modern technology that a country with

    insubstantial access to them will be able to develop and use very little modern technology.Indian industry has not felt their absence too much because we do not generate enough

    data in India, says Vijay Chandru, CEO of the Bangalore-based Strand Life Sciences.When this situation changes rapidly, as is already beginning to happen, Indian companies

    would desperately scrounge for supercomputing resources. You could always buy off-the-

    shelf supercomputers IBM, for example, has sold supercomputers in India but thesecomputers may not always come with a set of libraries and applications that are tailormade

    for Indian needs. It is unlikely that a commercial supercomputer would be used for nuclear

    simulations. Buying computers would not raise domestic supercomputing technologycapabilities either. Which is why the brains trust of the Indian supercomputing community

    suggests a judicious mixture of imports and domestic development.Apart from a national programme, some scientists are now trying to get Indian scientists

    hooked on to supercomputers around the world. Spenta Wadia, distinguished professor at

    the Tata Institute of Fundamental Research (TIFR) in Mumbai, is coordinating an effort to

    hook up Bangalore institutions to an international computing network called Gloriad. Thegovernmentled project would involve networked supercomputers at a large scale. The

    proposals to SAC-PM envisages the development of 5-6 machines working at 4-10 petaflops(thousand trillion computations per second) 20 machines at 100-200 teraflops (trillion

    computations per second), 50-60 machines at 1-10 teraflops.

    Indias fastest machine till recently, the Eka -- developed by the Tata ComputationalResearch Laboratories (CRL) in Pune --works at roughly 132 teraflops. The Eka was the

    fourth fastest computer in the world in 2007, but has fallen to 58th place now. In Indiaitself, it was overtaken by the computer built by the Isro, by combining a graphics processor

    unit from NVidia with a conventional central processing unit.

    This approach has been tried by the Chinese as well, but it works well for only a certainclass of problems. Isro needs accurate weather forecasts to plan its launches. Such

    accuracies, however, are increasingly necessary for other problems as well.

    Page 2 of 3Teraflop Turf: Bringing Back Indias Supercomp

    1/30/2012http://epaper.timesofindia.com/APA26300/PrintArt.asp?SkinFolder=pastissues2

  • 8/13/2019 Bringing Back Indias Supercomp

    3/3

    Page 3 of 3Teraflop Turf: Bringing Back Indias Supercomp

    1/30/2012http://epaper timesofindia com/APA26300/PrintArt asp?SkinFolder=pastissues2