Logo

HPC @ Uni.lu

High Performance Computing in Luxembourg

Gaia and Chaos End of Life Announcement

After 12 years of service for the Chaos cluster, and 8 years for the Gaia cluster, both systems have reached end of life. More than 6,2 million jobs were processed during that time, cumulating 13,8 MILLENIUM of CPU Time usage.
Throughout 2019 the workload and data have been migrated to our flagship Iris cluster, enabling now the final step and official decommissioning of these older systems.

Their access servers are now closed, and user access is not possible anymore.

On the 6th of January, the last of the active old hardware will be powered off and later sent to be recycled. Note that all hard disks will be wiped and destroyed.

History

Chaos was the first HPC cluster of the University, set up in 2007 by Dr. Sébastien Varrette in Kirchberg. Starting with 18 dual-core computing nodes and a backend storage of 3TB, it has been extended over time to reach a final capacity of 14.5TFlops with 81 nodes (1120 cores) in 2012.

Gaia had been set up in 2011 in Belval’s new BioTech I data centre and was initially composed of 72 Bull B500 compute nodes. It has been worked on by a growing HPC Team led by Dr. Varrette and composed of Hyacinthe Cartiaux, Fotis Georgatos, Valentin Plugaru and Sarah Peter. Gaia received upgrades until 2016 to reach a final capacity of 145 TFlops provided by 273 heterogeneous nodes (3440 cores), out of which 76 TFlops provided by GPU accelerators.

The HPC Team has maintained more than 1.8PB of live “Tier 0” (scratch) and “Tier 1” (home+project) storage on both clusters, ensuring high performance access, data security and integrity since the beginning.

During their lifetime, Gaia and Chaos processed respectively 4.5 million and 1.7 million jobs, cumulating 13835 Years of CPU Time usage.

Gaia and Chaos have each been used by over 740 and respectively 450 users comprising students, staff and project partners, for education and research on a large variety of projects.

The very first UL HPC School held in 2014 used both systems to train the 66 participants in the use of HPC platforms and scientific applications running on HPC.

RIP Chaos and Gaia clusters!
Yet sadness will be quickly replaced in 2020 with excitment as we are looking forward to University’s next HPC systems (codename: Aion, DLC-based and providing 1.7 PFlops of computing capacity) able to provide improved capabilities for research and teaching. More details to come…