Chalmers University of Technology is using Lenovo and Nvidia’s technology infrastructure to power its large-scale computer resource, or supercomputer, Alvis.
The project has seen the delivery and implementation of a clustered computing system for artificial intelligence and machine learning research, in what is Lenovo’s largest high-performance computing cluster for AI and ML in the Europe, Middle East and Africa region.
Alvis – old Norse meaning “all-wise” or “all-knowing” – is a national supercomputer resource within the Swedish National Infrastructure for Computing (SNIC).
It began initially in 2020 and has since developed to hold a capacity that solves larger research tasks on a broader scale. Financed by the Knut and Alice Wallenberg Foundation, the computer system is supplied by Lenovo and located at Chalmers University of Technology in Gothenburg, home to the EU’s largest research initiative, Graphene Flagship.
The collaborative project allows any Swedish researcher who needs to improve their mathematical calculations and models to take advantage of Alvis’ services through SNIC’s application system, regardless of the research field.
This supports researchers who are already utilising machine learning to analyse complex problems, and those who are investigating the use of machine learning to solve issues within their respective field, with the potential to lead to ground-breaking academic research in fields such as quantum computing and data-driven research for healthcare and science.
Noam Rosen, EMEA director, HPC and AI at Lenovo Infrastructure Solutions Group, says: “The Alvis project is a prime example of the role of supercomputing in helping to solve humanity’s greatest challenges, and Lenovo is both proud and excited to be selected as part of it.
“Supported by Lenovo’s performance leading technology, Alvis will power research and use machine learning across many diverse areas with a major impact on societal development, including environmental research and the development of pharmaceuticals.
“This computing resource is truly unique, built on the premise of architectures for different AI and machine learning workloads with sustainability in mind, helping to save energy and reduce carbon emissions by using our pioneering warm water-cooling technology.”
Sverker Holmgren, director of Chalmers e-Infrastructure Commons, hosting the Alvis system, says: “The first pilot resource for Alvis has already been used by more than 150 research projects across key Swedish universities.
“By making it much larger, and fully opening the Alvis Systems to all Swedish researchers, Chalmers and Lenovo are playing an important role in providing a national HPC ecosystem for future research.”
Chalmers has chosen to implement a scalable cluster with a variety of Lenovo ThinkSystem servers to deliver the right mix of NVIDIA GPUs to its users in a way that prioritises energy savings and workload balance.
This includes the Lenovo ThinkSystem SD650-N V2 to deliver the power of Nvidia A100 Tensor Core GPUs, and the Nvidia-Certified ThinkSystem SR670 V2 for Nvidia A40 and T4 GPUs.
Rod Evans, EMEA director of high-performance computing at Nvidia, says: “The work we’re doing with Chalmers University and its Alvis national supercomputer will give researchers the power they need to simulate and predict our world.
“Together, we’re giving the scientific community tools to solve the world’s greatest supercomputing challenges – from forecasting weather to drug discovery.”
The storage architecture delivers a new Ceph solution with 7.8 petabytes, to be integrated into the existing storage environment at Chalmers.
Nvidia Quantum 200 Gb/s InfiniBand provides the system with low- latency, high data throughput networking and smart in-computing acceleration engines.
With these high-speed infrastructure capabilities, users have almost 1000 GPUs, mainly Nvidia A100 Tensor Core, including over 260,000 processing cores and over 800 TFLOPS of compute power to drive a faster time to answer in their research.
In addition, Alvis leverages Lenovo’s Neptune liquid-cooling technologies to deliver unparalleled compute efficiency. Initially, full air cooling was proposed for the project, but Chalmers instead decided to deploy Lenovo Neptune warm water-cooling capabilities to reduce long-term operational costs and result in a “greener” AI infrastructure system.
As a result, the university anticipates that there will be significant energy savings thanks to efficiencies through water cooling.
The HPC solution has been in production since February 2022 and will be fully operational by Summer 2022.