During the ongoing RAISE 2020, the Government of India outlined that an AI-driven economy can radically transform our economy. Leveraging AI for inclusive growth is one of the core principles identified and a path for much-needed job creation in various sectors, apart from generating new business opportunities. According to experts, a nationwide artificial intelligence implementation can only be achieved with a specialised infrastructure, both hardware and software.
When it comes to robust exascale infrastructure, the Centre for Development of Advanced Computing under MeitY has been tasked to develop a nationwide grid of high-performance computing systems to facilitate the AI stack.
High-Performance Computing (HPC) is one of the leading technologies which can expedite the process of analysing data, data mining and simulation of all kinds of data like bioinformatics or computer vision models. The computation has to be divided into sub-tasks, and tens of thousands of processors are utilised to accelerate the analysis/simulations. Most of the challenges in sectors like CV or bioinformatics domains need infrastructure beyond the teraflop level of scale. The team at C-DAC is meeting these HPC requirements by porting a complete set of tools, databases, libraries, and hardware resources onto big supercomputing servers.
Subscribe to our Newsletter
Join our editors every weekday evening as they steer you through the most significant news of the day, introduce you to fresh perspectives, and provide unexpected moments of joy
C-DAC was formed in 1987, formerly as the Centre for Development of Advanced Computing Technology (C-DACT). C-DAC embarked on its first HPC mission in 1988. Since the beginning, it has produced a series of supercomputing systems known as PARAM series of supercomputers. Going all the way from PARAM 8000, India’s 1st Giga-scale supercomputer in 1990 to its first Teraflop supercomputer PARAM Padma in 2002, which was India’s first supercomputer to enter the Top500 list of supercomputers of the world.
C-DAC is a very critical part of AI missions for the country. The supercomputing systems and facilities of C-DAC are used to solve computationally challenging problems in a number of areas of practical significance.
“HPC and AI have converged together because whatever the computing needs which was required by AI were not present earlier and that’s now becoming a reality. Storage and compute are coming together has made a lot of things possible,” said Dr Hemant Darbari, Director General, C-DAC during his talk at the RAISE 2020 event.
HPC plays a critical role in the National Supercomputing Mission and in turn, builds the infrastructure for artificial intelligence C-DAC is working on India’s national grid infrastructure of HPC systems by bringing together academic and research institutions across numerous cities in India as part of Nation Knowledge Network (NKN).
“Under NKN, we have created a platform which will be made available to the startups, industry researchers and faculty members across India and take AI on a faster track,” said Dr Darbari.
C-DAC Is Creating First Of Its Kind National AI Infrastructure
One of the systems which C-DAC have created recently for handling incredibly large scale AI workloads includes Param- Siddhi AI, a supercomputer with 100 AI Petaflops and 2.5 million cores.
C-DAC has been given the task of building exascale capable systems as the key enabler for accelerating scientific discovery and innovation that will have an enormous societal return on investment. The objective is to establish a dependable and secure exascale ecosystem for AI workloads with innovative designs and human resources.
“Our goal is to develop hardware encompassing exascale chip design, and manufacture of exascale server boards, exascale interconnects, storage with embedded silicon photonics. To develop solutions for grand challenge problems for the national level,” spoke Dr Darbari.
C-DAC has established three facilities already at IIT BHU, IISER Pune and IIT Kharagpur. The endeavour involved training more than 1800 supercomputing manpower/faculties. Other systems which are going to be deployed across the country will provide around 75 institutions that will close work through NKN- the backbone for supercomputing systems.
These systems are already being used for use cases such as oil exploration, flood prediction and development of a multi-sectoral simulation lab to address urban environmental issues. Then there is NSM Platform For Genomics and Drug Discovery which is leveraging C-DAC infrastructure.
Apart from this, C-DAC has been working on developing exascale server board, interconnect, processor, system software libraries, storage and HPC-AI converged accelerator. All these things are being developed domestically in India as part of the National Supercomputing Mission.
Dr Darbari said, “We have developed and manufactured Indigenous servers (Rudra), which meet the HPC requirements of all governments and PSUs. This is the first time that a server system is being made in India along with the full software stack developed by C-DAC.”
Rudra is an indigenous server which has been developed and manufactured in India for meeting the compute requirements of government departments. Along with Rudra, CDAC gas developed HPC Interconnect Trinetra as well as a complete HPC system software stack for government HPC systems. It has also started developing an indigenous HPC processor.
Another concept which C-DAC has brought is ready to use supercomputing in a box. These are affordable solutions for academic, scientific and research institutions that are on the verge of adopting HPC, Deep Learning, AI and for skill development. More than 150 systems have been deployed. Such systems are called the Param Shavak, with 3-17 Teraflops leveraging GPUs from NVIDIA.
There are four kinds of Param Shavak systems:
- Param Shavak HPC System
- Param Shavak DL GPU (For Deep Learning)
- Param Shavak VR System (For AR/VR)
- Param Shavak Shrishti- GenNext Bio-Informatics Appliance (For Bio & Agriculture)
The objective of Param Shavak machine PARAM Shavak is a solution which can provide computational resources (with advanced technologies to do high-end computations for scientific, academic and engineering workloads involving modelling, simulation and data analytics. This initiative is hoped to produce HPC aware, skilled workforce and for promoting research by combining leading-edge emerging technologies at the grass-root level. Because the complexity of computational requirements continues to expand at colleges and universities, those places are looking for affordable solutions for advanced computing and developing models that can ingest large amounts of data at an affordable cost.