HPC is a strategic technology with high commercial yield

“A lot of people in these committees have not experienced this ecosystem around HPC. I attended one meeting in Bengaluru, where Intel was trying to sell its new chip. Before that, I attended one meeting in Goa, which was dominated by the industry. It all sounded very good but no one really had an idea what NSM actually wanted to do,” says an academic, who has since distanced himself from these meetings.

And the gerontocracy continues to throw its weight around.

From a soft power to a superpower

In 2011, when the Chinese machine Tianhe-1 became the fastest supercomputer in the world, the political class in Delhi felt that India was being left behind. Between 2012 and 2013, at the behest of the erstwhile Planning Commission, an elaborate plan was put together for an NSM, which incidentally had a projected cost of Rs 4,500 crore. But the UPA-II never got around to formally approving it. In March 2015, when the NDA government approved it, the language smacked more of catch-up nationalism than of a scientific mission. “As far as supercomputing is concerned, India is ranked at number 74 and China is number 1. There are 500 supercomputers in the world, and India has only 9,” said the minister at that time.

If China is number one, and it has been so for seven years in a row, it’s because it has consistently invested in building an HPC community ever since it started in the 1950s. By 1986, when China announced its famous ‘863’ programme to gain parity with the US and the rest of world in supercomputing, it was a coordinated effort to master all the related technologies in semiconductor manufacturing, the design of integrated circuits, mining and refining of rare earth metals, and other things. In 2016, China led the Top 500 list with 169 machines. The US came second, with 165 machines.

Recent policies in those two countries have been aimed at boosting supercomputing through internal investments and export controls. National security and domestic industries are cited as the top reasons. In July 2015, President Barack Obama issued an executive order authorising a National Strategic Computing Initiative.

Apart from the bragging rights and muscular power that come with supercomputers, the commercial uses for HPC are increasing by the day.

“We understood that the new kid on the block is biology. The largest requirements of the future are likely to be in biology, material science, finance and national data repositories,” says Narayanaswamy Balakrishnan, former chairman of the Supercomputer Education and Research Centre at the IISc and one of the architects of NSM. He wouldn’t comment on the current progress because he is given to believe that he is “no longer wanted”.

India’s supercomputing effort began in the last 1980s after it was denied the Cray supercomputer under a technology embargo, which led to the setting up of the Centre for Development of Advanced Computing (C-DAC). In 1990, when PARAM, a multiprocessor machine, was unveiled by the C-DAC, at 5 gigaflops, it was the second fastest supercomputer in the world at that time.

“After one mission in the 1980s, the whole of the 1990s got wasted because nobody consolidated the early gains. It became too tactical and routine,” says S Ramakrishnan, former C-DAC director general, who retired in 2009. “Once a government commits to such programmes, it must go on progressively escalating the budget, delivery and use. It shouldn’t become an accountant,” he said. By the 2000s, the technology controls were lifted, users could buy from wherever they wanted and India’s build programme weakened. “One or two people who had clout in Delhi and in purchase committees, they always favoured machines from multinational companies. In the land of the blind, one-eyed person is the king,” he added*.

The NSM hot potato

What Ramakrishnan does not admit is that the users and builders of HPC have never really communicated with each other in India. “It has grown over the years because people who have built these machines have not built anything that the users can use,” says a physicist from a government institution who has not been even remotely consulted under NSM.

(A list of the top Indian machines is here.)

The muddle continues. The appointment of Rajat Moona, who as director general of the C-DAC was a key professional in the NSM, ends this month, and he is off to a new assignment, as the director of IIT-Bhilai. Several emails and phone calls to his office in Pune remained unanswered. In January 2016, when the DST and MeITY had released Rs 110 crore towards NSM, Moona had said, “What is interesting is that this time the focus isn’t restricted to building the supercomputers but also creating applications that would benefit from it.”

 

Leave a Comment