▲An aerial view of GIST’s HPC-AI data center

- HPC-AI public infrastructure 6PFLPOS GPU, 10PByte storage, etc. will be supplied to GIST Super Computing Center.                                                                          

- A comprehensive infrastructure services such as AI design, construction, consulting, and operation will be provided through the CJ OliveNetworks-Kakao Enterprise-Dawon Computing Consortium.

- GIST will provide customized AI computing services to industry-academia partnership by establishing HPC-AI public infrastructure.

[CJ OliveNetworks=January 20, 2022 (Thursday)] CJ OliveNetworks (CEO Cha In-hyeok) won an order for building a high-performance computing-based artificial intelligence (HPC-AI) public infrastructure run by GIST (Gwangju Institute of Science and Technology, President Kim Ki-seon) and will contribute to vitalizing the industry-academia partnership.

Last year, GIST was selected as the managing organization for the HPC-AI public infrastructure construction project hosted by the Artificial Intelligence Industry Convergence Project Group as part of the artificial intelligence-centered industrial convergence cluster construction project. The HPC-AI public infrastructure will be built at the Super Computing Center on the GIST campus with the goal of opening in the first half of this year.

High Performance Computing (HPC) refers to the use of supercomputers or computer clusters to solve advanced computational problems. In today's world where AI technology leads to national competitiveness, competition is accelerating among countries, such as building hyperscale HPC-AI with powerful computational functions for AI research.

CJ OliveNetworks has experience in carrying out a number of large IT projects and has competitive advantage in the AI ??field, such as having developed AI solutions and retaining AI experts. This order was highly evaluated for cooperating in the form of a consortium with companies with the highest competitiveness in each field, such as CJ OliveNetworks, Kakao Enterprise, and Dawon Computing..

In particular, in order to efficiently respond to the ever-growing demands for AI performance, CJ OliveNetworks will carry out a system integration for GPU clusters and high-performance storage using good software quality certifications (GS certification) such as 'Kakao i Cloud' of Kakao Enterprise and AI platform Aprica's 'Cheetah'. Through this, it is expected to contribute to the localization of AI solutions and lead the domestic AI ecosystem.

CJ OliveNetworks will provide GIST with hyperscale comprehensive AI infrastructure services such as AI infrastructure design and construction, consulting, and operation through a total of three steps by the year 2023.

GIST's HPC-AI public infrastructure implements a super-large AI computing infrastructure, including 6PFLOPS (petaflops, a computer performance unit capable of processing 1,000 trillion operations per second) GPU with powerful computational capacity and 10PByte (petabyte) storage.

In addition, it plans to supply development resources for HPC-AI users such as companies, researchers, and students and provide an optimal AI research environment with an integrated system that includes △service portal, △platform service, and △dataset service.

The HPC-AI public infrastructure is equipped with AI as a service (AlaaS) allowing users to conveniently use everything from hardware to finished software that is required for AI research or application activities, and the price efficiency of paying only for what is used can be expected.

After establishing the HPC-AI public infrastructure, GIST plans to support AI computing for research and education and provide customized HPC-AI computing services for domestic industry-academia research institutes and global partners who need large-scale learning, in line with the national ultra-high performance computing innovation strategy, to lead the public infrastructure for the realization of AI.

“Due to the COVID-19 pandemic, the use of HPC is becoming more important in major countries, such as utilizing HPC’s powerful computational functions to develop therapeutics and vaccines,” said Seo Kyung-ho, Head of CJ OliveNetworks’ Infrastructure DT Business Team. “We will contribute to vitalizing the AI ??ecosystem by combining the competencies and know-how of the CJ OliveNetworks consortium to build an HPC-AI public infrastructure that combines HPC and AI.”

[Note] Petaflop

It refers to processing 1,000 trillion mathematical operations per second and is also known as PetaFlop or PFLOP. It is a new word created by combining the prefix Peta, which stands for the 15^th power of 10 in the International System of Units (SI) and the computer performance unit, Flops, which refers to the number of floating-point operations that can be performed per second. A computer equipped with a 1 petaflops processor has a processing speed that is 100 million times faster than a Pentium 133Mz processor.

[Note] Petabyte

It is a unit of capacity for digital information, representing (23)(byte) x (210)(kilobyte) x (210)(megabyte) x (210)(gigabyte) x (210)(terabyte) x (210) of a bit, the smallest unit of information processed in a computer. The word is made by combining the prefix Peta, which stands for the 15^th power of 10 and the unit used to indicate data in a computer, Byte, and it refers to the amount of data. 1PB is about 1 million GB, which is the capacity that is capable of holding 174,000 DVD movies (about 6GB).