Indian Supercomputing: Progress Awaited

India’s position among supercomputing nations have been on the wane, despite huge use cases for supercomputing in India, thanks to its large population. Despite policy announcements, progress has still been slow.

That India lags behind in the wider application of high-performance computing (HPC) is not exactly news.

But despite less-than-encouraging average, India’s best in many areas—from education to wealth creation to technology—have been comparable with the best in the world. And so was it for supercomputers, not long back.

India was home to the 4th fastest supercomputer in the world, in November 20017, according to the bi-annual Top500.org list that tracks and ranks the world’s fastest supercomputers. India’s EKA – a cluster platform-based supercomputer at the Computational Research Laboratories at Tata Sons, Pune—ranked behind just three faster supercomputers in the world—two in the United States and one in Germany. There was another at 58th position in that list.

India’s best was certainly among the world’s best.

Just for the record, India had nine supercomputers in the Top 500. China had 10, even though China’s fastest was slower than India’s second fastest.

Since then, with every single six-monthly ranking, China has increased its presence in Top 500. India had remained in that bracket—between two to eleven.

In the latest edition of the ranking, released in November 2019, China has 228—the highest in the world; yes, well ahead of the United States’ 199. India has two—both in the top 100, if that is any consolation.

The rise of China in supercomputing has been anything but an accident. There has been a result-oriented national strategy and a sustained policy that has delivered. The growth has been, by and large, gradual, even though there have been ups and downs in between, as in any resilient growth.

In India too, there have been policy discussions, attentions at the top level, and typical policy interventions such as a supercomputing mission. So far, it has resulted neither in creating a best nor in increasing the number. In fact, if anything, from 2015 onwards, there has been a gradual slide in the number of India supercomputers that have been featuring in the Top 500.

 

India’s top-level policy initiatives

The first high-profile policy stance, specifically focused on supercomputing came in September 2012, when Kapil Sibal was the minister for information technology in the UPA government. Sibal wrote to the then Prime Minister Dr Manmohan Singh, detailing a plan prepared by the state-owned research & development outfit, Centre for Advanced Computing (C-DAC) to build the fastest supercomputer by 2017, at a cost of some Rs 4,700 crores.

That is all we heard about this plan. Through 2013, very little policy decision happened. In May 2014, new NDA government took over and the prime minister announced the Digital India initiative. Though much of it was bringing together of the already planned initiatives, a big push such as this, directly from the prime minister, gave it a lot of push. A lot of technology-leveraged new plans (and some old plans, redesigned with added layers of digital technologies) meant that computing would get a further boost.

Getting more than a billion people to do things digitally, online, by itself means requirement of high-performance computing in massive scale. That is the reason why the use case for HPC or supercomputing is natural to China and India. But while China had realized that, India faltered. And we have seen the results.

However, with NDA government’s new plans, there was hope. What made one hopeful was previous research-led initiatives—including Sibal’s dream of building the fastest supercomputer in the world—were siloed research plans, isolated from applications. Building the fastest was more of a prestige issue. It was envisaged as a trophy.

This writer, in his personal blog, pointed it out then. “The focus is purely on speed. It is a petaflops speed supercomputer that the minister and C-DAC want to build. The application is secondary. While performance is not a bad objective to have, spending Rs 4700 crore to just be on top of the table sounds a little too much,” the blog said.

With Modi’s Digital India, any such plan would have a ready-made use case.

Before Digital India was launched, in December 2014, the Minister for Communications & IT, Ravi Shankar Prasad, had launched a C-DAC-built supercomputer, that was dubbed the world’s ‘most compacting supercomputer’ at that time by the government. This HPC initiative, called PARAM Shavak aimed to provide computations on a larger scale for the scientific, engineering and academic programs. C-DAC continued with its PARAM nomenclature, which it started in the 90s.

 

National Supercomputing Mission

In March 2015, the government launched the National Supercomputing Mission (NSM). The aim was to connect national academic and R&D institutions with a grid of over 70 high-performance computing facilities at an estimated cost of Rs 4,500 crore over seven years.

"We are going to install 73 supercomputers in different parts of the country, and all will be linked by a computer grid. This is about Rs 4,500 crore scheme in which Rs 2,800 crore will come from the Ministry of Science and Technology and the rest, about Rs 1,700 crore, from my department, IT," Communications and IT Minister Ravi Shankar Prasad said.

The mission statement said the supercomputers would also be networked on the National Supercomputing grid over the National Knowledge Network (NKN), a high-speed network that connects academic institutions and R&D labs.

“Academic and R&D institutions as well as key user departments/ministries would participate by using these facilities and develop applications of national relevance. The Mission also includes development of highly professional High Performance Computing (HPC) aware human resource for meeting challenges of development of these applications. The Mission implementation would bring supercomputing within the reach of the large Scientific & Technology community in the country and enable the country with a capacity of solving multi-disciplinary grand challenge problems,” the statement said.

That was a coherent plan and it gained credibility after the prime minister endorsed it publicly. “Supercomputer will do super-computing and will be the reason for super commitment,” PM Modi tweeted in July that year.

Since then, C-DAC has created application-oriented groups and has been working to create supercomputers for different applications.

In November 2018, C-DAC awarded a three-year contract to Atos to supply the latter’s BullSequana supercomputers.

Atos will deploy “its BullSequana supercomputers, with a cumulative computing power of more than 10 petaflops. These supercomputers, including the recently announced BullSequana XH2000 will be installed in various academic and research institutions,” said a statement from Atos.  

In February 2019, the Prime Minister inaugurated C-DAC-built supercomputer, PARAM Shivay, at Indian Institute of Technology, BHU, Varanasi. It has a peak computing power of 837 teraflops. The fastest supercomputer in the world is 200795 teraflops and the 500th in the Top 500 list is 2278 teraflops. The Indian fastest, Pratyush, at Indian Institute of Tropical Meteorology, has a Peak computing power of 4006 teraflops.  

According to the government statement, apart from this, two more supercomputers, designed, manufactured and assembled in India are being installed, one each at Indian Institute of Science, Education & Research (IISER) at Pune and IIT Kharagpur.

 

Measuring the Progress

While the plans of this government are far more coherent as compared to the one-off burst of the previous government, the progress has been less than satisfactory.

The National Supercomputing Mission was supposed to install 73 supercomputers in educational and R&D institutions across the country. In the first five years (almost) of the seven-year timeframe specified by NSM, only one supercomputer has been installed at IIT BHU and two are in pipeline.

Even if the installations may take a little more time, there is hardly any news beyond the three institutions, mentioned above. The NSM mission statement clearly said it would be ‘implemented’ in next seven years. In March 2020, it would complete five.

The progress has been less than satisfactory. We have seen what China has done in ten years.

What is also of concern is—even though C-DAC is working on various applications—there has been little news on application areas—beyond research and meteorology.  A research institution, however good it is, cannot work on applications effectively without active involvement of users themselves.

In the global competitiveness ranking, India has been faring badly in the technology readiness index. That is primarily due to lesser access to communication services, where India and China were once comparable.

Supercomputing is another area where one has seen China going past India decisively. We do not have to compete with China, but both the countries are similar in the sense we have huge population and any data concerning them would be huge and processing them would require huge computing power. They need supercomputers like no one else. The Dragon has jumped away; the tiger is still limping.

 


Add new comment