Supercomputer at UKSW supports science and economy

Challenges:

• enable the university to open up to new research directions;
• enable the university to carry out projects based on scientific achievements, using the latest artificial intelligence technologies, both for businesses and state and scientific institutions;
• implementation of cutting-edge software-defined storage, the first such SDS solution from IBM implemented in Poland.

Benefits:

• high-performance computing implementation, extended with artificial intelligence and high-performance data analysis;
• simulations, phenomena, and processes modeling implementation based on large datasets;
• a new research perspective for polish science;
• with machine learning and artificial intelligence, computers can perform the same tasks much faster and often more efficiently;
• optimization and forecasting applications, including, but not limited to, in medicine, industry, or finance. The computing cluster will accelerate scientists’ work and open up new opportunities to put high computing power into practice, from mathematical modeling to diagnostic imaging, to finding innovations for industry.

Solution components:

• IBM Power System servers with NVIDIA Tesla coprocessors;
• computing nodes connected to the high-performance, software-defined IBM Elastic Storage Server via the high-throughput InfiniBand network;
• a powerful training platform that guarantees high throughput when executing workloads generated by intensive computing, deep learning, and AI mechanisms.

A new research perspective for polish science

The Center’s goal is to open the university to new research directions and develop cooperation with business, especially using large-scale data processing and artificial intelligence achievements.

“Our reality is changing, therefore science must change as well. Our research center project will primarily be based on scientific advances, using the latest artificial intelligence technologies, both for business and state and scientific institutions,” says Professor Marek Niezgodka, director of the Center for Digital Science and Technology at UKSW. “Its very creation is an example of a model cooperation between the science, represented by our University, and the economy.”

The innovative hardware architecture was developed and built by IBM.

“We leveraged the technologies used in IBM Power System servers equipped with NVIDIA Tesla coprocessors. We connected the computing nodes to the efficient, software-defined IBM Elastic Storage Server storage using the high-throughput InfiniBand network”, says Krzysztof Lipski, IT Engineer at Infonet Projekt, responsible for the implementation. “Based on this, we designed and deployed a powerful training platform that guarantees high throughput when executing workloads generated by intensive computing, deep learning, and AI mechanisms,” he adds.

The computer that was named CATO is a coherent system that encompasses both a large-scale data processing subsystem and a data storage and sharing subsystem. The system is complemented by software packages that operate in the specific layers of the system.

“The implementation of this project shows that we not only perfectly understand and navigate technological issues, but also effectively connect this world with the world of science and business,” points out Piotr Fabianski, President of the Management Board of Infonet Projekt. “The implementation of cutting-edge software-defined storage is the first such SDS solution from IBM implemented in Poland. We are proud to have an impact on technological development,” he adds.

Innovation development catalyst backed by AI

The computing cluster launched is GPGPU accelerated. The state-of-the-art hardware architectures used, combined with the software, create an efficient working environment. The areas where CATO can be applied naturally result from the wide variety of applications of artificial neural networks in information processing. The neural networks’ self-learning abilities and their similarities to the characteristics of the human brain translate into domain-specific applications that are sure to grow and expand. It is at the stage of innovation that the question of so-called deep learning, which is a rapidly growing field of computer science, becomes particularly relevant.

“By building our solution based on the latest hardware architectures and using an ecosystem of software solutions immersed in the open source community, we want to act as an innovation catalyst,” emphasizes Michal Iwanski from UKSW CNT. “With our implementation we want the world of supercomputers that used to be hermetic and intended for the most experienced specialists implementation to become accessible and effective to use. We don’t want to change scientists’ habits of using traditional tools, but to move them into a new dimension, into the computing cluster space, together with their languages, libraries, and tools,” Iwanski concludes.