Architecture

This diagram presents a high-level overview of data flow in NativeLink.

1. The client such as Bazel, Buck2 or Reclient creates a job and sends it to the scheduler’s job queue.
2. The scheduler finds a suitable worker in the worker pool and routes the job to it.
3. The worker runs the job, sending output artifacts to the cas.
4. The worker provides download instructions for the artifact to the scheduler.
5. The scheduler forwards the download instructions to the client.

In conclusion, the client created a job and sent it to NativeLink and gets a response that lets it download the built artifact.