compute next to your data

Having your data well-managed in the cloud is only half the job. The people and applications that need it — analysts in QGIS, a dashboard in the browser, a downstream model, a customer paying for a data product — each expect a different protocol, and none of them want to download a petabyte first.

So every organization builds a delivery tier: tile servers, subsetting APIs, caches, autoscaling, authentication. It is expensive to build, and it never stops needing attention.

• Duplicated effort — a separate service to build and operate for every protocol you need to speak.

• Data movement — copies extracted and staged for each consumer, immediately going stale.

• Scaling pain — request spikes on an operational feed become an on-call problem.

Compute is the Earthmover platform's data delivery service. It runs immediately next to your data and serves any Arraylake repository over the protocols your consumers already speak. There is no pipeline to write and no infrastructure to provision: you enable a service, and every repository, branch, tag, and commit becomes addressable over HTTP.

Because requests are served from the data's own storage, only the result crosses the network — a tile, a timeseries, an aggregate — instead of the archive it was computed from. Access inherits the permissions already on the repository, so an operational feed and a private research dataset are served by the same machinery under different rules.

# Enable a service on your organization
al compute enable my-org tiles

# Every repo, branch, tag, and commit is addressable
https://compute.earthmover.io/v1/services/tiles/
  {org}/{repo}/{branch|tag|commit}/tiles/WebMercatorQuad/{z}/{y}/{x}
    ?variables=temperature_2m
    &colorscalerange=-30,40