What is Ray?
How Ray runs distributed AI workloads in TauGrid
Core technologies
Understand the systems TauGrid brings together
TauGrid brings several focused systems into one workflow. Each one owns a different part of the path from submission to saved results.
| Technology | Role in TauGrid |
|---|---|
| Ray | Runs distributed Python, training, tuning, and serving workloads |
| Kueue | Decides when a workload can use shared cluster quota |
| adx-mon | Optionally sends cluster and experiment telemetry to Azure Data Explorer |
| Stellar | Helps teams inspect and compare experiment evidence |
Developers mainly interact with TauGrid and Stellar. Platform teams operate Kueue, KubeRay, and optional adx-mon integrations.
What is Kueue?
How Kueue shares quota and admits TauGrid workloads
What is adx-mon?
How adx-mon sends Kubernetes and experiment telemetry to Azure Data Explorer
What is Stellar?
How Stellar turns saved run data into experiment views and comparisons