Hardware and Network Requirements

There are minimum node hardware and network requirements necessary for installation and testing of the Fawkes system.

Hardware requirements

The Fawkes hypervisor has the following requirements and recommendations for installation and testing.

Hypervisor Node Requirements

Hypervisor nodes are home to virtual machines, these require hardware-assisted virtualization and will be home to the management and Kubernetes VMs.

Table 1. Hypervisor Nodes
Type Requirements and Recommendations

CPU

x86_64.[1] Hardware-assisted virtualization is required. 8-core processor minimum for testing; 16-core or above is recommended for production.

Memory

8 GB minimum for testing; 32 GB or above recommended for production.

Disk Capacity

100 GB minimum for testing; 1 TB or above recommended for production.

Disk Quantity

1 for testing; 3 or more recommended for production, 2 of which will be used for OS redundancy. The third disk should be larger than 1.5 TB for ephemeral, virtual machine storage.

Disk Performance

5,000+ random IOPS per disk(SSD/NVMe). Management nodes (first 3 nodes) must be fast enough for Etcd.

Network Card

1 Gbps Ethernet minimum for testing, at least two 25 Gbps Ethernet for production.

Network Switch

Trunking of ports required for VLAN support (IEEE 802.1Q tagging).

Storage Node Requirements

Storage nodes play a pivotal role in the deployment of an HPC cluster. These nodes are not deployed as hypervisors and do not run any virtualization services other than Podman.

The current storage solution of choice is CEPH.

Table 2. Storage Nodes
Type Requirements and Recommendations

CPU

x86_64.[2]. 8-core processor minimum for testing; 16-core or above is recommended for production.

Memory

8 GB minimum for testing; 32 GB or above recommended for production.

Disk Capacity

100 GB minimum for testing; 4 TB or above recommended for production.

Disk Quantity

1 for testing; 8 or more recommended for production, 2 of which will be used for OS redundancy. The remaining disks will be used for the storage solution.

Disk Performance

5,000+ random IOPS per disk(SSD/NVMe). Management nodes (first 3 nodes) must be fast enough for Etcd.

Network Card

1 Gbps Ethernet minimum for testing, at least four 25 Gbps Ethernet for production.

Network Switch

Trunking of ports required for VLAN support (IEEE 802.1Q tagging).


1. hypervisor aarch64 artifacts are not built at this time.
2. aarch64 support is not planned for storage nodes.