The University of Ljubljana’s Faculty of Computer and Information Science in Slovenia is set to introduce a state-of-the-art modular data center to advance its AI research efforts. This containerized facility will be positioned atop the faculty’s building, offering powerful computational resources for groundbreaking projects like the development of a Slovenian Large Language Model (LLM).
A Strategic Investment in AI Infrastructure
The modular data center, supplied by NTR Inženiring, comes with a hefty price tag of €2.5 million ($2.57 million). Of this investment, €870,000 ($895,930) will be funded by Slovenia’s Research and Innovation Agency, while the remaining balance will be provided by the faculty itself.
Enhancing Research with Cutting-Edge Supercomputing Capabilities
The new data center is designed to significantly boost the faculty’s AI research capabilities. With its integration of classical and supercomputing technologies, the facility will provide on-site access to a high-performance supercomputer. This setup will allow researchers to conduct complex calculations and efficiently train deep neural networks, a vital component of AI model development.
“With the new data center, our faculty will gain on-site access to a supercomputer, enabling our researchers to perform calculations and train deep neural network models more efficiently,” said Dean Mojca Ciglarič.
The PoVeJMo Project: A Leap Forward in LLM Development
The data center will play a crucial role in the faculty’s PoVeJMo (Let’s say it) project, an initiative to develop an open-access Slovenian LLM. The model will be trained using vast amounts of diverse text data, representing a significant milestone in Slovenia’s AI ambitions.
The integration of advanced technologies like air and direct liquid cooling will also ensure that the data center achieves high energy density, supporting the sustainable operations of the supercomputer.
NTR Inženiring’s Expertise in Supercomputing Solutions
NTR Inženiring, the company behind the modular data center’s supply, brings considerable experience to the project. Previously, NTR played a key role in the development of Slovenia’s Vega and Maister supercomputers. NTR’s director, Primož Mahorič, highlighted the exceptional energy density and technological advancements of the new facility: “This facility integrates various classical and supercomputing technologies while achieving exceptionally high energy density through the combined use of air and direct liquid cooling.”
Slovenia’s Data Center Landscape and Future Growth
While Slovenia has a relatively small data center market, Ljubljana hosts the majority of the nation’s facilities. Key operators in the sector include Telemach, Datacenter.si, Perftech, and SoftNet. As demand for high-performance computing grows, particularly in AI and deep learning fields, the role of advanced data centers like the one at the University of Ljubljana will continue to be crucial.
The Vega supercomputer, inaugurated in 2021, remains one of the country’s flagship supercomputing resources, with a performance of 6.9 petaflops. This new modular data center promises to complement these efforts by offering more accessible computational power for research and development.
FAQ Section
1. What is the purpose of the University of Ljubljana’s new modular data center?
The data center will support advanced AI research, particularly in training deep neural networks and developing a Slovenian LLM.
2. How much did the modular data center at the University of Ljubljana cost?
The project is valued at €2.5 million ($2.57 million), with €870,000 ($895,930) funded by Slovenia’s Research and Innovation Agency.
3. What cooling technologies are used in the modular data center?
The facility will use a combination of air and direct liquid cooling to achieve high energy density.
4. Who is supplying the modular data center?
The data center is being supplied by NTR Inženiring, a company with experience in supercomputing infrastructure.
5. How will the new data center contribute to AI research?
It will provide on-site access to a supercomputer, enabling more efficient AI model training and computational research.
