Supercomputing in India has a history going back to the 1980s.[1] The Government of India created an indigenous development programme as they had difficulty purchasing foreign supercomputers.[1] As of June 2023[update], the AIRAWAT supercomputer is the fastest supercomputer in India, having been ranked 75th fastest in the world in the TOP500 supercomputer list.[2] AIRAWAT has been installed at the Centre for Development of Advanced Computing (C-DAC) in Pune.[3]
India had faced difficulties in the 1980s when trying to purchase supercomputers for academic and weather forecasting purposes.[1] In 1986 the National Aerospace Laboratories (NAL) started the Flosolver project to develop a computer for computational fluid dynamics and aerospace engineering.[4][5] The Flosolver MK1, described as a parallel processing system, started operations in December 1986.[4][6][5]
In 1987 the Indian Government had requested to purchase a Cray X-MP supercomputer; this request was denied by the United States government as the machine could have a dual use in weapons development.[7] After this problem, in the same year, the Government of India decided to promote an indigenous supercomputer development programme.[8][9][10] Multiple projects were commissioned from different groups including the Centre for Development of Advanced Computing (C-DAC), the Centre for Development of Telematics (C-DOT), the National Aerospace Laboratories (NAL), the Bhabha Atomic Research Centre (BARC), and the Advanced Numerical Research and Analysis Group (ANURAG).[9][10] C-DOT created "CHIPPS": the C-DOT High-Performance Parallel Processing System. NAL had started to develop the Flosolver in 1986.[4][11] BARC created the Anupam series of supercomputers. ANURAG created the PACE series of supercomputers.[10]
The Centre for Development of Advanced Computing (C-DAC) was created at some point between November 1987 and August 1988.[8][10][9] C-DAC was given an initial 3 year budget of Rs375 million to create a 1000MFLOPS (1GFLOPS) supercomputer by 1991.[10] C-DAC unveiled the PARAM 8000 supercomputer in 1991.[1] This was followed by the PARAM 8600 in 1992/1993.[10][9] These machines demonstrated Indian technological prowess to the world and led to export success.[10][9] Param 8000 was replicated and installed at ICAD Moscow in 1991 with Russian collaboration.
The PARAM 8000 was considered a success for C-DAC in delivering a gigaFLOPS range parallel computer.[10] From 1992 C-DAC undertook its "Second Mission" to deliver a 100 GFLOPS range computer by 1997/1998.[1] The plan was to allow the computer to scale to 1 teraFLOPS.[10][12] In 1993 the PARAM 9000 series of supercomputers was released, which had a peak computing power of 5 GFLOPS.[1] In 1998 the PARAM 10000 was released; this had a sustained performance of 38 GFLOPS on the LINPACK benchmark.[1]
The C-DAC's third mission was to develop a teraFLOPS range computer.[1] The PARAM Padma was delivered in December 2002.[1] This was the first Indian supercomputer to feature on a list of the world's fastest supercomputers, in June 2003.[1]
By the early 2000s it was noted that only ANURAG, BARC, C-DAC and NAL were continuing development of their supercomputers.[6] NAL's Flosolver had 4 subsequent machines built in its series.[6] At the same time ANURAG continued to develop PACE, primarily based on SPARC processors.[6]
The Indian Government has proposed to commit US$2.5 billion to supercomputing research during the 12th Five-Year Plan period (2012–2017). The project will be handled by Indian Institute of Science (IISc), Bangalore.[13] Additionally, it was later revealed that India plans to develop a supercomputer with processing power in the exaflops range.[14] It will be developed by C-DAC within the subsequent five years of approval.[15]
Supercomputing overview | |
---|---|
Formed | 2015 |
Parent department | C-DAC |
Website | https://nsmindia.in/ |
In 2015 the Ministry of Electronics and Information Technology announced a "National Supercomputing Mission" (NSM) to install 73 indigenous supercomputers throughout the country by 2022.[16][17][18][19] This is a seven-year program worth $730 million (Rs. 4,500 crore).[citation needed] Whilst previously computer were assembled in India, the NSM aims to produce the components within the country.[20] The NSM is being implemented by C-DAC and the Indian Institute of Science.[19]
The aim is to create a cluster of geographically distributed high-performance computing centers linked over a high-speed network, connecting various academic and research institutions across India.[17] This has been dubbed the "National Knowledge Network" (NKN).[20] The mission involves both capacity and capability machines and includes standing up three petascale supercomputers.[21][22]
The first phase involved deployment of supercomputers which have 60% Indian components.[19] The second phase machines are intended to have an Indian designed processor,[19] with a completion date of April 2021.[20] The third and final phase intends to deploy fully indigenous supercomputers,[19] with an aimed speed of 45 petaFLOPS within the NKN.[20]
By October 2020, the first assembled in India supercomputer had been installed.[20] The NSM hopes to have the manufacturing capability for indigenous production by December 2020.[20]
As of November 2023[update] there are 4 systems based in India on the TOP500 supercomputer list.[23]
Rank | Site | Name | Rmax (TFlop/s) |
Rpeak (PFlop/s) |
---|---|---|---|---|
75 | Centre for Development of Advanced Computing | AIRAWAT - PSAI[24][25] | 8.5 | 13.17 |
163 | Centre for Development of Advanced Computing | PARAM Siddhi-AI | 4.62.0 | 5.27 |
201 | Indian Institute of Tropical Meteorology | Pratyush (Cray XC40)[26][27] | 3.76 | 4.01 |
354 | National Centre for Medium Range Weather Forecasting | Mihir (Cray XC40) | 2.57 | 2.81 |
List | Number of systems in TOP500 |
System Share (%) | Total Rmax (Gflops) |
Total Rpeak (Gflops) |
Cores |
---|---|---|---|---|---|
2020 June | 2 | 0.4 | 6,334,340 | 6,814,886 | 202,824 |
2019 November | 2 | 0.4 | 6,334,340 | 6,814,886 | 202,824 |
2019 June | 3 | 0.6 | 7,457,490 | 8,228,006 | 241,224 |
2018 November | 4 | 0.8 | 8,358,996 | 9,472,166 | 272,328 |
2018 June | 5 | 1 | 9,078,216 | 10,262,899 | 310,344 |
2017 November | 4 | 0.8 | 2,794,753 | 3,759,153 | 107,544 |
2017 June | 4 | 0.8 | 2,703,926 | 3,935,693 | 103,116 |
2016 November | 5 | 1 | 3,092,368 | 4,456,051 | 133,172 |
2016 June | 9 | 1.8 | 4,406,352 | 5,901,043 | 204,052 |
2015 November | 11 | 2.2 | 4,933,698 | 6,662,387 | 236,692 |
2015 June | 11 | 2.2 | 4,597,998 | 5,887,007 | 226,652 |
2014 November | 9 | 1.8 | 3,137,692 | 3,912,187 | 184,124 |
2014 June | 9 | 1.8 | 2,898,745 | 3,521,915 | 169,324 |
2013 November | 12 | 2.4 | 3,040,297 | 3,812,719 | 188,252 |
2013 June | 11 | 2.2 | 2,690,461 | 3,517,536 | 173,580 |
2012 November | 9 | 1.8 | 1,291,739 | 1,890,914 | 90,548 |
2012 June | 5 | 1 | 787,652 | 1,242,746 | 56,460 |
2011 November | 2 | 0.4 | 187,910 | 242,995 | 18,128 |
2011 June | 2 | 0.4 | 187,910 | 242,995 | 18,128 |
2010 November | 4 | 0.8 | 257,243 | 333,005 | 25,808 |
2010 June | 5 | 1 | 283,380 | 384,593 | 30,104 |
2009 November | 3 | 0.6 | 199,257 | 279,702 | 23,416 |
2009 June | 6 | 1.2 | 247,285 | 333,519 | 33,456 |
2008 November | 8 | 1.6 | 259,394 | 368,501 | 37,488 |
2008 June | 6 | 1.2 | 189,854 | 275,617 | 32,432 |
2007 November | 9 | 1.8 | 194,524 | 303,651 | 34,932 |
2007 June | 8 | 1.6 | 45,697 | 86,642 | 10,336 |
2006 November | 10 | 2 | 34,162 | 61,520 | 10,908 |
2006 June | 11 | 2.2 | 36,839 | 66,776 | 11,638 |
2005 November | 4 | 0.8 | 11,379 | 21,691 | 3,354 |
2005 June | 8 | 1.6 | 13,995 | 24,726 | 4,212 |
2004 November | 7 | 1.4 | 6,945 | 11,873 | 2,126 |
2004 June | 6 | 1.2 | 5,652 | 9,557 | 1,750 |
2003 November | 3 | 0.6 | 2,099 | 5,098 | 1,106 |
2003 June | 2 | 0.4 | 1,158 | 3,747 | 822 |
India began developing supercomputers in the late 1980s after being refused one by the US.
L.M. Patnaik developed a significant amount of the factual material for this report.