Supercomputing in India

[porky_kicker]

What happened to the exaflop supercomputer development program ?

 

[HariPrasad-1]

What is the status of our exa flop computer which was about to get ready between 2017 to 2020?

@Indx TechStyle do you have any idea? I am curious.

 

[Indx TechStyle]

What is the status of our exa flop computer which was about to get ready between 2017 to 2020?

@Indx TechStyle do you have any idea? I am curious.

Unknown

 

[Rxbanda]

It would be very helpful if academic researchers can get access to these systems. Like the Anton of DE Shaw research.

 

[Kumata]

Got it. But u r comparing the memory & storage.

So I said it to be a bad idea.

Yeah.. thats what the core of any computer is along side processor.. they are offering a decade old scalable processor series with this so called jugaad.. i mean computing world have moved on ffrom this to next level...

I can get a quad socket / 6 TB rack mount these days.. Make a grid off it and can do what ever i want.. more so with 40 Gig becomig the defacto standard for network at core of DC's......off lately storage is so cheap that NVMe is new defacto standard in Ent DC's ... even flash is dead in Ent DC's .. magnetic disks are a history already...

 

[Karthi]

What is the status of our exa flop computer which was about to get ready between 2017 to 2020?

@Indx TechStyle do you have any idea? I am curious.

Atos and C-DAC sign Cooperation Agreement for global advancement in Quantum Computing, Artificial Intelligence and Exascale Computing in India

C-DAC (Centre for Development of Advanced Computing), a national premier R&D organization under Ministry of Electronics and Information Technology, Government of India, and Atos , a global leader in digital transformation, today announce that they have signed a Cooperation Agreement for...

atos.net

CDAC said it will be ready at 2022. ,US, Japan, Taiwan and China also have similar time lines

 

[HariPrasad-1]

Atos and C-DAC sign Cooperation Agreement for global advancement in Quantum Computing, Artificial Intelligence and Exascale Computing in India

C-DAC (Centre for Development of Advanced Computing), a national premier R&D organization under Ministry of Electronics and Information Technology, Government of India, and Atos , a global leader in digital transformation, today announce that they have signed a Cooperation Agreement for...

atos.net

CDAC said it will be ready at 2022. ,US, Japan, Taiwan and China also have similar time lines

Thanks brother. Please enlighten me more about this I am very curious.

 

[Karthi]

Thanks brother. Please enlighten me more about this I am very curious.

Don't know much Brother , CDAC made progress in designing an EXA scale Mechine . May be they are trying to develop own processor for super computers . These machines are energy hungry and eat electricity in 50-100 MW . Cooling requirements are far more than what an average super Computer want . Don't know the progress in these areas . Meanwhile China developed a prototype of ExaScale machine in 2018 .

 

[LETHALFORCE]

Japan's Fugaku retakes supercomputing crown from US and China

Riken-Fujitsu system clocks in nearly three times faster than top IBM machine

asia.nikkei.com

Japan retakes supercomputing crown from US and China

 

[ezsasa]

India progressing rapidly towards the goal of indigenously made #Supercomputers. National Super Computing Mission (NSM) is boosting high power computing in the country

 

[ezsasa]

India progressing rapidly towards the goal of indigenously made Supercomputers

National Super Computing Mission (NSM) is boosting high power computing in the country

India is fast expanding its supercomputer facilities and developing the capacity to manufacture its own supercomputers in the country.

The National Super Computing Mission (NSM) is rapidly boosting high power computing in the country through its various phases to meet the increasing computational demands of academia, researchers, MSMEs, and startups in areas like oil exploration, flood prediction as also genomics, and drug discovery.

With the infrastructure planned in NSM Phase-I already installed and much of Phase-II in place, the network of supercomputers through the country will soon reach to around 16 Petaflops (PF). Phase-III, to be initiated in January 2021, will take the computing speed to around 45 Petaflops.

NSM is jointly steered by the Ministry of Electronics and IT (MeitY) and Department of Science and Technology (DST) and implemented by the Centre for Development of Advanced Computing (C-DAC), Pune and the Indian Institute of Science (IISc), Bengaluru.

Param Shivay, the first supercomputer assembled indigenously, was installed in IIT (BHU), followed by Param Shakti and Param Brahma at IIT-Kharagpur and IISER, Pune, respectively.

Thereafter supercomputing facilities were set up in two more institutions, and one is being set up in Phase-I, ramping up high power computing speed to 6.6 PF under Phase-1. In Phase-II, 8 more institutions will be equipped with supercomputing facilities by April 2021, with a total of 10 PF compute capacity. MoUs have been signed with total 14 premier institutions of India for establishing Supercomputing Infrastructure with Assembly and Manufacturing in India. These include IITs, NITs, National Labs, and IISERs. Some of these have already been installed, and some more will be done by December this year. The Phase-II installations will be completed by April 2021.

Work on Phase-III will start in 2021 and will include three systems of 3 PF each and one system of 20PF as a national facility.

The three phases will provide access to High-Performance Computing (HPC) Facilities to around 75 institutions and more than thousands of active researchers, academicians working through Nation Knowledge Network (NKN) - the backbone for supercomputing systems.

HPC and Artificial Intelligence (AI) have converged together. A 100 AI PF Artificial Intelligence supercomputing system is being created and installed in C-DAC, which can handle incredibly large-scale AI workloads increasing the speed of computing-related to AI several times.

The mission has also created the next generation of supercomputer experts by training more than 2400 supercomputing manpower and faculties till date.

Indigenous capability

Powered by the NSM, India’s network of research institutions, in collaboration with the industry, is scaling up the technology and manufacturing capability to make more and more parts in India. While in Phase-I, 30 percent value addition is done in India, that has been scaled up to 40 percent in Phase-II.

Efforts are being made to design and develop parts like server board, interconnect, processor, system software libraries, storage, and HPC-AI converged accelerator domestically.

India has developed an Indigenous server (Rudra), which can meet the HPC requirements of all governments and PSUs. This is the first time that a server system was made in India, along with the full software stack developed by C-DAC.

Experts said that the pace at which things are moving forward, we may soon have the motherboards and sub-systems manufactured in India, making the supercomputers indigenously designed and manufactured.

Such indigenously designed systems with most parts designed and manufactured in India will be installed at places like IIT-Mumbai, IIT-Chennai, and Inter-University Accelerator Centre (IUAC) at Delhi, C-DAC, Pune, which are covered under Phase-III and help move towards supercomputers developed and manufactured totally in India paving the way for self-reliance in the field.

India progressing rapidly towards the goal of indigenously made Supercomputers

India is fast expanding its supercomputer facilities and developing the capacity to manufacture its

pib.gov.in

 

[Indx TechStyle]

India’s ‘Param Siddhi’ AI supercomputer ranked at 63 out of 500 top most powerful supercomputers in the world
This will accelerate R&D in the war against COVID-19 through faster simulations, medical imaging, genome sequencing and forecasting and is a boon for Indian masses and for start-ups and MSMEs in particular.

The supercomputer with Rpeak of 5.267 Petaflops and 4.6 Petaflops Rmax (Sustained) was conceived by C-DAC and developed jointly withsupport of Department of Science and Technology (DST), Ministry of Electronics and Information Technology (MeitY) under NSM. (Representational Image: iStock)

Param Siddhi, the high-performance computing-artificial intelligence (HPC-AI) supercomputer established under National Supercomputing Mission (NSM) at C-DAC, has achieved a global ranking of 63 in TOP 500 most powerful non-distributed computer systems in the world released on 16 November 2020.

The AI system will strengthen application development of packages in areas such as advanced materials, computational chemistry & astrophysics, and several packages being developed under the mission on the platform for drug design and preventive health care system, flood forecasting package for flood-prone metro cities like Mumbai, Delhi, Chennai, Patna and Guwahati.

This will accelerate R&D in the war against COVID-19 through faster simulations, medical imaging, genome sequencing and forecasting and is a boon for Indian masses and for start-ups and MSMEs in particular.

It is a boon for application developers and will help testing of weather forecasting packages by NCMRWF & IITM, geo-exploration packages for oil and gas recovery; packages for aero-design studies; computational physics and mathematical applications and even online courses for education.

The supercomputer with Rpeak of 5.267 Petaflops and 4.6 Petaflops Rmax (Sustained) was conceived by C-DAC and developed jointly with the support of Department of Science and Technology (DST), Ministry of Electronics and Information Technology (MeitY) under NSM.

“It is a historical first. India today has one of the largest supercomputer infrastructures in the world and that is evidenced by the ranking that Param Siddhi-AI has received today,” said Secretary Department of Science & Technology, Professor Ashutosh Sharma.

“I truly believe that Param Siddhi-AI will go a long way in empowering our national academic and R&D institutions as well as industries and start-ups spread over the country networked on the national supercomputer grid over the National Knowledge Network (NKN),” Prof Sharma added.

He pointed out that with the infusion of Param Siddhi-AI, the scientific and technology community in the country will further be enabled and empowered to solve multidisciplinary grand challenges of healthcare, agriculture, education, energy, cybersecurity, space, AI applications, weather and climate modelling, and urban planning to name a few.

“This is a compelling piece of our journey in Atmanirbharta through Science Technology & Innovation,” he stressed.

Param Siddhi Supercomputer is built on the NVidia DGX SuperPOD reference architecture networking along with C-DAC’s indigenously developed HPC-AI engine, software frameworks and cloud platform and will help deep learning, visual computing, virtual reality, accelerated computing, as well as graphics virtualization.

 

[Indx TechStyle]

So, we are now capable of making PetaFlop computers on our own. Will break into top 10 too near future then.

 

[shade]

So, we are now capable of making PetaFlop computers on our own. Will break into top 10 too near future then.

Hardware is from Nvidia.
Software package(s) developed by C-DAC in collaboration with come French company called Atos.

I am in the dark however about what this "HPC-AI Engine" is, it is a hardware system/product? is it software?.
Same press release is repeated in a lot of news sites.

 

[Indx TechStyle]

Hardware is from Nvidia.
Software package(s) developed by C-DAC in collaboration with come French company called Atos.

Former problem will remain till a manufacturing becomes normal, latter has become an expertise now gained.

I am in the dark however about what this "HPC-AI Engine" is, it is a hardware system/product? is it software?

Looks like an AI software-hardware set.

 

[shade]

The Param Siddhi system made up of 42 of these modules

NVIDIA DGX Platform

The best of NVIDIA AI - all in one place.

www.nvidia.com

Each contains the below

• 8X NVIDIA A100 GPUS WITH UP TO 640 GB TOTAL GPU MEMORY
  12 NVLinks/GPU, 600 GB/s GPU-to-GPU Bi-directonal Bandwidth

• 2 6X NVIDIA NVSWITCHES
  4.8 TB/s Bi-directional Bandwidth, 2X More than Previous Generation NVSwitch

• 3 10X MELLANOX CONNECTX-6 200 Gb/s NETWORK INTERFACE
  500 GB/s Peak Bi-directional Bandwidth

• 4 DUAL 64-CORE AMD CPUs AND 2 TB SYSTEM MEMORY
  3.2X More Cores to Power the Most Intensive AI Jobs

• 5 30 TB GEN4 NVME SSD
  50 GB/s Peak Bandwidth, 2X Faster than Gen3 NVME SSDs

Per the white paper describing these modules in great details here, it is specifies that the CPUs are AMD Epyc 7742, which is a 64 core/ 128 thread 3.4Ghz CPU, there are 4x of these in each module

https://www.nvidia.com/content/dam/en-zz/Solutions/Data-Center/dgx-a100/dgxa100-system-architecture-white-paper.pdf

So the hardware is not indeginous per se.

 

[Indx TechStyle]

The Param Siddhi system made up of 42 of these modules

NVIDIA DGX Platform

The best of NVIDIA AI - all in one place.

www.nvidia.com

Each contains the below

• 8X NVIDIA A100 GPUS WITH UP TO 640 GB TOTAL GPU MEMORY
  12 NVLinks/GPU, 600 GB/s GPU-to-GPU Bi-directonal Bandwidth
• 2 6X NVIDIA NVSWITCHES
  4.8 TB/s Bi-directional Bandwidth, 2X More than Previous Generation NVSwitch
• 3 10X MELLANOX CONNECTX-6 200 Gb/s NETWORK INTERFACE
  500 GB/s Peak Bi-directional Bandwidth
• 4 DUAL 64-CORE AMD CPUs AND 2 TB SYSTEM MEMORY
  3.2X More Cores to Power the Most Intensive AI Jobs
• 5 30 TB GEN4 NVME SSD
  50 GB/s Peak Bandwidth, 2X Faster than Gen3 NVME SSDs

Per the white paper describing these modules in great details here, it is specifies that the CPUs are AMD Epyc 7742, which is a 64 core/ 128 thread 3.4Ghz CPU, there are 4x of these in each module

https://www.nvidia.com/content/dam/en-zz/Solutions/Data-Center/dgx-a100/dgxa100-system-architecture-white-paper.pdf

So the hardware is not indeginous per se.

Wondering if these would be license produced.

 

[Chinmoy]

Hardware is from Nvidia.
Software package(s) developed by C-DAC in collaboration with come French company called Atos.

I am in the dark however about what this "HPC-AI Engine" is, it is a hardware system/product? is it software?.
Same press release is repeated in a lot of news sites.

HPC-AI is both hardware and software based system where same algorithm runs parallel into different connected systems.
The designing of the whole system was done in house. So there is no point is crying over the hardware used as ALMOST all the SC of world use same hardware.

The AI engine, Software framework, Cloud platform are BTW developed by C-DAC.

 

[NAMICA]

I don't think any think any company in India manufacture such high end hardware ..It is our major weakness.

 

[SKC]

Hardware is from Nvidia.
Software package(s) developed by C-DAC in collaboration with come French company called Atos.

I am in the dark however about what this "HPC-AI Engine" is, it is a hardware system/product? is it software?.
Same press release is repeated in a lot of news sites.

Processors can be from outside, not a big issue. Even the first chinese supercomputer to reach rank 1 was made by using Intel processors which made US ban Intel from supplying processors to china for such kind of use.
The whole system should be designed in India with our own software package. Also, these supercomputers don't use general purpose OS. They need specialized OS to run according to the need.