New Delhi: A tender floated by the government for buying four supercomputers for its climate change programmes highlights the seriousness with which it is approaching the issue (of climate change) and the lack of it when it comes to developing supercomputers.
The machines the government plans to buy will be of 6-15 teraflop speed and will be put at the National Centre for Medium Range Weather Forecasting in New Delhi, the Indian National Centre for Ocean Information Services in Hyderabad, the Indian Institute of Tropical Meteorology and Indian Meteorological Department, both in Pune. A teraflop, or trillion floating point operations per second, is a measure of a computer’s speed. A floating point is a way of representing huge and complex numbers in simple ways using just two smaller numbers.
The cost of these machines isn’t known because supercomputers are built to order and not bought off the shelf. A 1 teraflop machine developed at The Institute of Mathematical Sciences in Chennai cost Rs2.5 crore.
“We had a pre-bid meeting last month, inviting leading supercomputer manufacturers. The technical bids will be opened at the end of this month, which will be followed by a separate financial bid later this year,” said a senior official of the earth sciences ministry, which looks after India’s climate change programmes.
The official added that supercomputer manufacturers such as IBM Corp., Hewlett Packard Sales India Pvt. Ltd, Cray Inc. and the government’s Centre for Development of Advanced Computing (C-DAC) attended the pre-bid meeting. None of the companies confirmed whether they had placed bids.
Given the government’s increasing focus on using robust, scientific climatic models to capture the effects of global warming, scientists say super-computers are vital to accurately develop future climate patterns. “The faster our computers, the better chance we have of simulating the present weather,” said Raman Sukumar, a member of the government’s expert committee on climate change. “The clearer our picture of today’s weather, the better our forecasting models,” he added.
Giant supercomputers such as the 280.6 IBM BlueGene/L housed at the Lawrence Livermore National Laboratories at the University of California, and Japan’s NEC Earth Simulator with a 35.8 teraflop speed, crunch weather models at international weather forecasting centres. The average speed of computers used for this purpose is around 25 teraflop; India’s fastest supercomputers do not go beyond 1 teraflop.
Ever since C-DAC developed Param Padma, the country’s fastest supercomputer in 2003, little progress has been made in the development of faster, cheaper computers. C-DAC claimed that it would launch a high-speed supercomputer in June, but this has been delayed, said a C-DAC official who did not wish to be identified. “This will be for bio-informatics applications and we are waiting for the right time to formally launch it,” he added. Param clocked a speed of 1 teraflop, giving it a ranking of 171 in Top500, a biannual supercomputer rating system developed by the universities of Mannheim in Germany and Tennessee in the US. As of 2007, a ranking in Top500 requires a minimum speed of 4 teraflops.
“US and Japan do not hesitate to fund supercomputing projects extensively since they are considered to be strategic to their future,” said Vijay Bhatkar, a former C-DAC director involved with manufacturing the Param supercomputers.
“In contrast, the Indian government has been following a flip-flop policy, with most of our R&D institutions not knowing what the user organizations need. Unless user organizations and R&D institutions collaborate, this disconnect will remain,” he added.
However, Narayanaswamy Balakrishnan, a computer sciences expert at IISc Bangalore and a member of the technical bid evaluation panel, said it was “meaningless to develop supercomputers just for the sake of it, because what could be done by supercomputers even a decade ago can be achieved by PC-based systems. Now given that we have identified areas like bio-informatics, weather-modelling as imperative to our needs, I am confident that we will be able to further our supercomputing capabilities”.