← All articles

Technical notes

Setting up an HPC cluster on AWS

This is an eight-part series on setting up an HPC cluster on AWS. The main design elements of the cluster will be as follows


HPC cluster architecture with storage, login, head and compute nodes, retaining all original network and service connections.
HPC cluster architectureView original ↗



The cluster will have seven virtual machines (VMs)

  • One head / control node

  • One login node

  • Three compute nodes

  • Two storage nodes

  • All the VMs will have the OS Rocky Linux 9.6 (Blue Onyx)

    • AMI: ami-0f2425d4cce4e97dd

    • Instance Type: t3.2xlarge

  • BeeGFS is used as the shared filesystem

  • OpenPBS is used as the HPC job scheduler

  • OpenLDAP is used for user management

  • Prometheus and Grafana is used for cluster monitoring


We use the following nomenclature throughout

  • Each machine in the cluster will be called node

  • The laptop or the desktop we use will be called local system


The cluster setup is divided into eight parts


The AMI and instance type used here may not be available on the AWS free tier. The main GitHub repo for is available here.