OPF-Ray runs inside AI Studio and relies on a small number of core platform components.
| Component | Purpose | What Users See |
|---|---|---|
| AI Studio Console and Notebook | Entry point for creating clusters and running code | UI and Jupyter-based workflow |
| Cluster Management Service | Validates requests and converts cluster specifications into deployable resources | Create, list, and delete cluster operations |
| KubeRay-backed Ray Cluster | Provides the actual distributed compute runtime | Head node, workers, client endpoint, dashboard |
| Teradata Connectivity | Lets workloads read and write enterprise data | SQL driver and parallel Arrow Flight SQL access |
| External Persistent Storage | Stores checkpoints, artifacts, and spill data where needed | AWS S3, object store, or shared file store depending on environment |
At a high level, you sign in to AI Studio, request a Ray cluster, wait for it to become ready, connect from a notebook, run distributed workloads, monitor progress, and delete the cluster when finished.