Render farm

loading…

 

Occupancy

 

How this pipeline works →  ·  The scheduler: what it is doing, what waits, where it is stuck →

Serving

Tokens served by the text models

Videora inference — monitored, not render capacity

Pushed from the Dubai workstation every 15 s; this page re-reads it every 15 s. Bhilai is polled once a minute over SSH, because it answers customer traffic and every poll is a round trip. Read-only — nothing here can start, stop or change a job.
If the timestamp stops advancing, the workstation has stopped publishing, not necessarily stopped rendering.