India’s Eka, which once figured among the world’s top 10, falls to 129 in global rankings; China takes lead in Asia
By Leslie D’Monte
In November 2007, India’s Eka was ranked the world’s fourth-largest supercomputer. Built by the Computational Research Laboratories Ltd (CRL), a unit of Tata Sons, the Hewlett-Packard (HP) system cost $30 million (around Rs 165 crore) and was built in just six weeks.
It was the first time that an Indian supercomputer figured among the world’s top 10.
The country had nine supercomputers in that Top 500 list, with two of them being in the Top 100—the other being an IBM system from the Bangalore-based Indian Institute of Science (IISc). China, on the other hand, had 10 supercomputers in the November 2007 list but just one, ranked 59, in the Top 100 list.
Four-and-a-half years later, the tables have turned dramatically.
India is falling far behind in the supercomputer race. So what’s ailing the country’s supercomputing abilities? Our technology editor, Leslie D’ Monte, finds out
According to the June 2012 Top 500 Supercomputers list (it is published twice in a year), Eka is now ranked 129. Two supercomputers figure in the top 100—the CSIR Centre for Mathematical Modelling and Computer Simulation (C-MMACS), ranked 58, and SAGA-220, developed by the Indian Space Research Organisation (Isro) and the IISc to solve complex aerospace problems, which is ranked 86. China, on the other hand, has raced ahead with 68 supercomputers, of which seven figure in the Top 100, with an NUDT supercomputer from the National Supercomputing Center in Tianjin ranked 5. A US supercomputer named Sequoia—an IBM BlueGene/Q system installed at the Department of Energy’s Lawrence Livermore National Laboratory—tops the list. Fujitsu’s K Computer installed at the RIKEN Advanced Institute for Computational Science (AICS) in Kobe, Japan, is now the No. 2 system. It held the No. 1 spot on the previous two lists.
Computational fluid dynamics model of the EKA data centre. The US is a clear leader in the high performance computing (HPC) segment with 253 of the 500 systems, followed by Europe with 107 systems. Dominant countries in Asia are China (68) and Japan with 34 systems.
Supercomputers, introduced in the 1960s and designed primarily by Seymour Cray at the Control Data Corp. (CDC), are used for calculation-intensive tasks such as problems relating to quantum physics, weather forecasting, climate research, oil and gas exploration, molecular modelling and physical simulations, detonation of nuclear weapons and research into nuclear fusion.
PARAM 8000 is considered India’s first supercomputer. It was built by the Centre for Development of Advanced Computing (C-DAC) with Russian collaboration. The C-MMACS Intel-powered systems will be used for a range of applications from climate-modelling to computational fluid dynamics (CFD) in aerospace. The project, reported Mint in March, will cost about Rs 50 crore. The National PARAM Supercomputing Facility (NPSF), which houses the PARAM Yuva, and the Bioinformatics Resources and Applications Facility (BRAF), which houses the Biogene and BioChrome supercomputing cluster, are used by researchers to solve problems in atmospheric sciences, nanosciences, strategic areas of engineering and research in cancer besides biomolecular drug research. But India sorely lacks speed in this area with the country’s supercomputers confined to teraflops (trillions of calculations per second). Getting to the petaflop level would mean increasing the processor power by 10 times and the cooling to 4,000 tonnes, besides the increase in floor space, all of which require a new architecture since growth in scale becomes non-linear at this stage, say experts.
SAGA-220 is powered by NVIDIA’s Tesla chips and Vishal Dhupar, managing director, sales and marketing, NVIDIA India, said for India to get into the big supercomputing league, it needed a “high performance computing ecosystem which involved hardware, storage capacity and scalable application software”. He explained that data centres and power are critical for supercomputers.
“A decent supercomputer would require 5-10 megawatts (MW) of power. Now that’s a problem in a country like India which faces a power shortage,” said Dhupar. He added that China had taken “bigger steps and hence has got the edge”.
“Supercomputer is a very niche area like car racing. Not all countries take it seriously all the time. Some like the US, Japan and Germany take it seriously all the time. China has taken it super serious. India, as usual, goes through fits and starts—CDAC Param once, Param Padma next, Eka suddenly,” said S. Sadagopan, director of the International Institute of Information Technology (IIIT), Bangalore. “Institute of Mathematical Sciences, in Chennai built one, C-MACCS is building a real super one. They plan to build one more by 2017—perhaps the most focused effort. In essence, we are doing okay, not great. It is also an expensive proposition.”
Dhupar also pointed out the importance of combining graphics processing units (GPUs) that are capable of handling the heavy computing tasks, while freeing the central processing units (CPUs) for other basic computing”.
“Only 60 of the top 500 supercomputers have GPUs and CPUs. This is bound to change in the coming years,” he said.
Meanwhile, the government needs to do much more if it wants India to compete in the big league, say experts. On 6 June, Ashwani Kumar, minister of state for planning, science and technology, said the government had allocated Rs 5,000 crore in the 12th Five-Year Plan to help India join the big league in supercomputing. He made the statement while visiting the National Chemical Laboratory in Pune.
