A group of identically configured worker Droplets within a DOKS cluster that run Kubernetes pods. A DOKS cluster can have multiple node pools, each with different Droplet sizes and auto-scaling configurations — enabling workload separation between CPU-intensive batch jobs and memory-hungry web services within the same cluster.
An ML platform uses two DOKS node pools: a general-purpose pool (s-4vcpu-8gb × 3) for API services and a GPU node pool (GPU Droplet × 2) for inference workloads. Kubernetes node selectors route inference pods exclusively to GPU nodes, while API pods scale horizontally across the general pool.