Fulhame

Fulhame is one of three Arm-based supercomputer systems belonging to the Catalyst UK programme.

Fulhame

The Catalyst UK programme is a collaboration between Hewlett Packard Enterprise, Arm, SUSE and three UK universities aiming to accelerate the adoption of Arm based supercomputer systems in the UK. EPCC was chosen as a site for one of these systems, Fulhame, with the other two systems being installed at the University of Bristol and the University of Leicester.

Fulhame was delivered and installed in late 2018. This HPE Apollo 70-based system consists of 64 compute-nodes, each with two 32-core Cavium ThunderX2 processors (ie 4096 cores in total), 128GB of memory composed of 16 DDR4 DIMMs, and Mellanox InfiniBand interconnects. 

It can be made available to both industry and academia, with the intention of building applications that drive economic growth and productivity as outlined in the UK Government’s Industrial Strategy.

Technology

# Nodes

64 Compute Nodes

CPUs

2x ARM Cavium ThunderX2, 1.6-2.5GHz, 32-core per node

Total CPU cores

4,096

Other key technology or accelerator details

Each Thunder X2 processor can operate in SMT-1, 2 or 4 mode giving each processor the option of up to 128 threads (256 per compute-node).

System Memory details

128GB per node for a total of 8TB of RAM

Storage technologies and specs

180TB Lustre hosted on HPE Apollo Cavium ThunderX2 servers

Interconnect technologies and specs

Mellanox EDR Infiniband at 100 Gbit/s

Any other tech points not captured

First production deployment of a fully homogenous ARM-based Lustre system.

Cooling tech and specs

The system is composed of 64 Apollo 70 air cooled servers contained in two racks along with management and storage systems.

Scheduler details

Slurm

System OS Details

SuSE Linux Enterprise Server (SLES) 15 sp 1

Access

Academic access is granted by application to EPCC stating proposed application usage, requested time, impacts of research. Contact EPCC in the first instance to discuss.

People

Fulhame is managed by Dr Michèle Weiland. The system is maintained by the HPC Systems Teams within EPCC and the hardware provider HPE in conjunction with the Catalyst project partners (Arm, SUSE, nVidia).