Cirrus

Cirrus is a state-of-the-art HPC system running industry-standard Linux that provides an ideal platform for users to solve their computational simulation and modelling challenges.

Users are able to run their own codes as well as accessing a range of commercial software tools. These are accessed on an on-demand, pay-per-use basis and used to solve CFD and FEA simulation and modelling problems in sectors such as:

• Automotive
• Aerospace
• Energy
• Oil and gas
• General engineering
• Life sciences
• Financial services

Technology

# Nodes
280 CPU, 36+2 GPU nodes

CPUs
CPU Nodes: 2xIntel Xeon “Broadwell”, 2.1 Ghz, 18-core per node
GPU Nodes: 2xIntel Xeon “Cascade Lake”, 2.4 Ghz, 20-core per node

Total CPU cores
10,080 Broadwell cores, 1520 Cascade Lake cores

GPUs
4xNvidia Tesla V100-SXM2-16GB GPU, 640 Tensor core, 5,120 CUDA core per node, 152 GPUs in total

Total GPU cores
97,280 Tensor cores, 778,240 CUDA cores

System Memory details
CPU Nodes: 256GB per node, 71 TB total
GPU Nodes: 384GB per node, 14 TB total

Storage technologies
1PB Ceph distributed file system for critical data storage.

1PB HPE E1000 ClusterStor Lustre parallel file system provides high performance data access with high capacity.

256TB HPE high performance solid state storage.

Interconnect technologies and specs
CPU Nodes, login nodes and Lustre: 26 FDR Inifinband Switches at 56 Gbit/s
GPU Nodes and NVMe storage: 8 EDR Infiniband Switches at 100 Gbit/s

Layout/Physical system scale
Cirrus is composed of 4 HPE SGI 8600 cabinets as well as 4 support cabinets for system management and storage and a Cooling Distribution Unit

Cooling tech and specs
The Cirrus Compute Nodes are stored in water cooled HPE SGI 8600 cabinets

Scheduler details
Slurm

System OS Details
CentOS 8 (Compute Nodes) and Red Hat Enterprise Linux (Login and management nodes)

Science and applications

Cirrus provides an ideal platform for solving computational, simulation, modelling, and data science challenges. Work on Cirrus covers a wide range of fields including physics, chemistry and engineering.

Access

Academic access

Academic access to Cirrus is primarily available for work within the EPSRC remit and academics at Scottish Institutions. Special arrangements for access to Cirrus are available for University of Edinburgh researchers.  

Details of academic access routes are available on the Cirrus website.

Please contact the service desk if you have any questions regarding access, support@cirrus.ac.uk 

Commercial access

Cirrus has been designed to support a wide range of industry applications and EPCC can provide both access to HPC resources and also consultancy to effectively exploit HPC resources.

For access to Cirrus and EPCC expertise please contact support@cirrus.ac.uk.

Trial access

Details of instant access are available on the Cirrus website.

People

The key management staff for Cirrus are:

• Prof. Mark Parsons (Service Director)
• Dr Alan Simpson (Deputy Service Director)
• Ms Josephine Beech-Brandt (Service Owner)

Support

System maintenance and provision are provided by the Service Desk, Computational Support and Engineering and HPC Systems Team within EPCC and the hardware provider HPE.

Support is available Monday to Friday from 08:00 until 18:00 UK time (excluding UK public holidays) via support@cirrus.ac.uk.