Node
A single computer in a cluster, usually with a GPU and some CPU cores.
Think of it like
One person on a team.
Example
A node: one server with 8 Nvidia H100 GPUs, 256GB CPU RAM, 2TB NVMe storage.
How it actually works
A node is the unit you allocate for jobs. Small nodes (1 GPU) are good for inference; large nodes (8+ GPUs) are good for training. You pay per node, per hour. Sharing nodes between jobs is complex (resource contention, debugging) so most clusters run one job per node.
For product teams
Node cost is your primary budget unit. Know how many you need.
For engineers
Understand your cluster's node config. Profile your job's resource usage per node.
Related
- Cluster building block.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome