New AI cluster available to scientists

Prague, 6 February 2020 A new computer cluster has been made available to scientists, researchers, and university students in the Czech Republic. It was built by the CESNET Association and named ADAN. The term ‘cluster’ denotes several computers connected together via a high-speed network. The ADAN cluster boasts exceptional computational power, significantly enhancing the existing CESNET portfolio of high-performance computers.

The name ADAN was not chosen by accident: in the Elvish language of J. R. R. Tolkien’s fantasy world it means ‘human’. The ADAN cluster’s specifications make it ideally suited for machine learning, a subset of artificial intelligence research.

‘The ADAN cluster is composed of 61 nodes, each of which is equipped with two NVIDIA Tesla T4 accelerator cards. Its Omni-Path high-speed, low-latency connection technology allows individual nodes to cooperate when solving complex or parallel tasks that exceed the capabilities of a single node,’ says Miroslav Ruda, the head of the Distributed Computing Department of the CESNET Association.

Costs for the ADAN cluster topped CZK 11 million. It was integrated into the CESNET Association’s MetaCentre, which is tasked with making available and coordinating the resources of participating institutions within a unified academic distributed computing infrastructure, part of the European EGI e-infrastructure. The MetaCentre’s capacity is currently 26,560 CPU cores.

The MetaCentre’s services are intended for scientific and research organisations as well as university students across the country. These services are put to use by experts in many fields, including high-energy particle physics, biotechnology, medicine, meteorology, and astronomy. The MetaCentre provides them with interconnected computational and data sources for solving tasks that are beyond the capabilities of any one workplace. ADAN brings MetaCentre’s potential to a new level.

Accessing the MetaCentre’s computational resources is easy. Applicants from the domestic academic community can find all relevant information at https://www.MetaCentre.cz/ and can start computing almost immediately. User support is also available to make the first steps in this new environment easier and to provide expert help with specific questions and requirements from research teams.

ADAN Cluster Configuration
• 61 nodes, each with 2 x 16-core Xeon Gold 5218 CPU @ 2.30 GHz, 192 GB RAM
• Each node has 2 x GP-GPU NVIDIA T4 accelerator cards
• Each node has local 800 GB SSD scratch filesystem
• Omni-Path high-speed internal network
• 1 gigabit ethernet in each node, 10 gigabit uplink into the CESNET core network
• CESNET e-infrastructure integration, allowing immediate access to other services, such as direct connection to CESNET data storage
• Singularity container solution technology allows the immediate use of preloaded software from the NVIDIA GPU cloud (https://ngc.nvidia.com/)

The project CESNET E-Infrastructure – Modernisation (CZ.02.1.01/0.0/0.0/16_013/0001797) is co-financed by the European Union.

The CESNET Association was founded by various Czech universities and the Czech Academy of Sciences in 1996. It is engaged in research and development in the field of information and communication technologies and builds and develops the CESNET national e-infrastructure for research and education. Thanks to its research activities and accomplishments, it represents the Czech Republic in important international projects, particularly in building the pan-European GÉANT data network and grid projects (EGI.eu), and actively participates in their implementation. The Association also works on the use of high-speed networks for sharing multimedia data both through synchronous video conferencing and shared applications as well as asynchronous streaming.
More information: www.cesnet.cz

Last change: 7.2.2020