Template and topology
Pin orchestrator version, accelerator SKU, node pools, interconnect, image, storage, labels, taints, and zones.
Loading AuraOne ComputePreparing GPU search and provider inventory.
Compute / GPU clusters
Choose an orchestrator, provider profile, region, accelerator topology, node-pool range, private-network policy, storage, access, and cleanup boundary.
Cluster creation requires authentication, an enabled provider profile, a current template, quota, budget approval, and an idempotent operation. This public page does not create infrastructure.
| Orchestrator | Typical fit | Required inspection |
|---|---|---|
| Kubernetes | Serving, training platforms, operators, and mixed services | Template, version, network, node-pool, and access controls |
| Slurm | Scheduled multi-node training and HPC-style batch workloads | Partition, queue, topology, storage, and scheduler policy |
| Ray | Distributed Python, training, tuning, and batch execution | Head/worker topology, autoscaling, images, and object storage |