Deploying a private registry that engineers can pull from vs. sending an image via docker load/save

Hey folks,

We’re building docker images on a dedicated instance and want to return the image back to our engineers’ machines as fast as possible. We’re currently running “docker save” in our cloud and “docker load” in our client, which has a pretty high end-to-end latency.

We’re wondering if things would be faster if we deployed the built image to a private registry in our compute, and had our clients run a “docker pull” on the image that they’ve built.

Would this be any faster? Are there other ways to optimize?