History

Prospero is a collective response to the growth of a number of independent HPC facilities for research within LJMU (particularly those of the Astrophysics Research Institute, the School of Engineering, and the School of Biological Sciences), each of which outgrew the capabilities of the local infrastructure for power and cooling. Migration to a consolidated, centralised solution hosted in the university’s primary data centre has enabled the development of a facility with capacity and capability significantly beyond those of department-specific facilities.

Initial system:

  • Development of a new system was catalysed by the award of two major research grants to the ARI in 2018/19, from the European Research Council (PI: Prof. Ian McCarthy) and the Royal Society (PI: Prof. Rob Crain), and enabling an ambitious facility to be planned.

  • HPE, in collaboration with UK vendors DTP, were approached to provide a design proposal meeting the broad requirements of the university’s research base.

  • An initial system comprising 20 standard compute nodes and a memory-rich compute node (for a total of 1344 compute cores and 11.5TB memory) saw `first light’ in August 2020.

  • The system was integrated with the ARI’s existing Lenovo 1.5PB parallel data store.

Major updates:

  • April 2021: funding from UKRI’s World Class Laboratories program enabled a first major system upgrade, adding 7 standard compute nodes, 250TB of NFS storage and a second gateway/login node for high availability. This brought Prospero to a total of 1792 compute cores and 15TB of memory.

  • May 2021: a second Lenovo 1.5PB parallel data store was added for use by the BAHAMAS simulations team (PI: Prof. Ian McCarthy).

  • April 2022: a new Royal Society grant awarded to the ARI (PI: Prof. Rob Crain) was matched by LJMU, enabling a futher 9 nodes to be added, bringing the system to 2368 compute cores and 19.5TB of memory.

  • February 2023: LJMU partnered with Alces Flight, to provide HPC systems administration and management services for Prospero.

  • May 2023: Prospero become a heterogeneous platform, incorporating its first accelerated node equipped with 4 Nvidia A100 GPUs. This node was funded by the Faculty of Engineering & Technology’s RCIF allocation, following the submission of a business case led by Dr. Ian Jarman and Dr. Ivan Olier-Caparroso.

  • June 2023: 8 standard compute nodes were added, funded by the ARI, bringing the system to a total of 2888 compute cores and 24.5TB of memory.

  • August 2023: 8 standard compute nodes were added, funded by LJMU, bringing Prospero to its present complement of 52 standard compute nodes, 1 memory-rich compute node and 1 accelerated node, for a total of 3400 compute cores, 4 GPU engines and 28.5TB of memory.

  • September 2023: a second memory-rich node has been ordered, enabled by a Royal Society grant awarded to the ARI (PI: Prof Rob Crain) and matched funding from the Faculty of Engineering & Technology’s RCIF allocation.