RACK SYSTEMS documentationRun jobs on racks you can hold

The API, the scheduler, the telemetry feed and the SLA behind C1 in us-west. Written for the people who will run the jobs.

Quickstart

Check capacity, submit a job and read its telemetry with curl or Python. Five minutes once you have a key.

Start here

Images, commands, storage mounts, plans and what happens when a batch job is preempted.

Read the guide

C1 hardware in full. Eight H100 SXM on one NVLink fabric, network, storage and commissioning.

See the spec

How reserved, on demand and batch hours are placed on the same rack, and who gives way first.

How placement works

One sample per GPU per second. Schema, retention and the websocket stream.

Read the schema

Three plans with one number each, what is billed, and what we owe you when C1 is unavailable.

See the numbers