Historically, India has struggled to establish an advantage in supercomputing—but its government is looking to change that with the announcement of a slew of new supercomputers and new, homegrown technology, including its “Rudra” server platform. India’s Ministry of Science & Technology announced this week that it is planning to deploy nine new supercomputers this year in a continuation of India’s National Supercomputing Mission, which has sought since 2015 to install a large number of supercomputers in the country with a steadily increasing percent of indigenous components.
The nine new supercomputers, the ministry said, will be commissioned and installed in the coming year in institutes like IIT (Indian Institute of Technology) Bombay, IIT Madras, IIT Patna, IIT Delhi, the Inter-University Accelerator Centre (IUAC), the Centre for Development of Advanced Computing (C-DAC) Pune, the S.N. Bose National Centre for Basic Sciences (SNBNCBS), the National Centre for Radio Astrophysics (NCRA) and the National Informatics Centre (NIC).
The news came below the fold in an announcement that predominantly sought to highlight current wins from the National Supercomputing Mission: namely, 10 institutions that have already installed supercomputing infrastructure under the plan. Those institutions include the Indian Institute of Science (IISc), IISER (Indian Institute of Science Education and Research) Pune, the Jawaharlal Nehru Centre for Advanced Scientific Research (JNCASR), the National Agri-Food Biotechnology Institute (NABI) and various C-DACs and IITs. The ministry added that “final stage installation work is being carried out in five more institutions.”
In total, it appears: 15 systems with nine more ostensibly on the way. The ministry says that this will provide hundreds of institutions and thousands of researchers with access to HPC facilities.
Still, this falls a bit short of the initial objectives of the National Supercomputing Mission, which was funded to the tune of $600 million USD and, when it was initiated in 2015, sought to deploy 73 supercomputers in India by 2022—roughly three times the new total anticipated within the coming year.
Currently, India has three public supercomputers powerful enough to rank on the Top500: the Atos-built PARAM Siddhi-AI, installed in 2018 at C-DAC Pune (4.6 Linpack petaflops, #102); and the HPE-built Pratyush and Mihir systems, both installed in 2018 at the Indian Institute of Tropical Meteorology (respectively, #121 at 3.8 Linpack petaflops and #228 at 2.6 Linpack petaflops).
The ministry also emphasized that design and development of indigenous server nodes, interconnect switches, storage technologies and software stacks for HPC is ongoing with “85 percent indigenous manufacturing,” suggesting that these technologies are intended for use in forthcoming supercomputers. By way of example, they highlighted India’s “first indigenous server platform” (called Rudra) and a “next-generation indigenous HPC interconnect” (called Trinetra). Rudra, detailed in December, is a dual-socket, Intel-based server that supports GPU acceleration (see graphic above).
The newfound emphasis India is placing on its National Supercomputing Mission and homegrown HPC design and manufacturing is yet another signal that, as Atos’ Eric Eppe put it, “the pandemic [has] reminded everyone that it was important to be sovereign.” Initiatives like Europe’s EuroHPC, European Processor Initiative and European Chips Act have increasingly sought to insource the elements of the supercomputing supply and operation chains that have, historically, been outsourced to global HPC leaders like the United States and China.