Skip to main content

Edge Clusters

Technical Details​

Edge clusters are Kubernetes clusters made up of one or more GPU-enabled Kubernetes Nodes.

Hardware​

VisualCortex supports most current generation Intel or AMD x64 processors. VisualCortex required at least 16GB of RAM for each node in the Edge Cluster. At least one of the Edge Cluster nodes needs to have an NVIDIA GPU with compute capability of at least 7.5. See the official NVIDIA Website for more information. Each node should have storage space of at least 250GB - preferably SSD. Significantly more storage is required if you intend to utilise the VisualCortex Investigations module.

VisualCortex also supports the NVIDIA Jetson Orin AGX for very small deployments - Contact Us if you intend to use one of these devices.

Software​

VisualCortex supports Ubuntu Server as its operating system. Both 22.04 LTS and 24.04 LTS are currently supported.

Networking​

In terms of networking, Edge Clusters need outbound network access to the Control Plane. This outbound access needs to support Web Sockets. Most customers elect to have outbound internet access too, but this is not required. No inbound access to the Edge Cluster's network is required - this means no port forwarding, no DMZs etc. VisualCortex Edge Clusters will also work behind NAT.

Management​

Edge Cluster management is an adminsitrative feature. You'll need to be either an organisation adminsitrator or super administrator to add, remove and modify edge clusters. See Persmissions for more information. You can see a list of Edge Devices by choosing the Edge management tile in the Configuration section.

Edge Administration

Settings​

Edge Clusters have a number of settings. These are outlined below:

  • Status - This can either be available or disabled. Disable an Edge Cluster if you don't want VisualCortex to use it
  • Name - This is the name which appears throughout VisualCortex
  • Description - An optional description
  • Physical Location - An optional text field representing where the Edge cluster is physically located
  • Physical GPUs - This is the total number of GPU cores present across the cluster
  • Usable GPUs - This is the total number of GPU cores you want used at any one time. This is typically set to a lower number that Physical GPUs when operating in a Highly Available cluster
  • Cluster Architecture - This is the architecture of the cluster itself.

Provisioning​

VisualCortex has an automated installer which simplifies the setup process for provisioning Edge Clusters. Upon creating a device, a button titled Generate Setup Command appears. Clicking that will give you a short-lived URL which you can paste or type into your physical Edge Cluster node.

Edge Setup

The command generated will spawn a Terminal-based GUI which will guide you through the setup process.

Edge Installer