Cloud-native formats

COG, Zarr, GeoParquet and STAC: reading the part you need over the network instead of downloading the file.

A range request pulling two megabytes out of a forty-gigabyte remote file instead of downloading it. read the part you need, over the networkscene.tif — 40 GBHTTP range request2 MBwhat crossed the wire40 GBthe old wayCOG, Zarr and GeoParquet are all the same idea: an internal layout a reader can seek into

A cloud-native format is an ordinary format with an internal layout a reader can seek into, plus a server that honours HTTP range requests. That combination changes the default move from "download the file, then filter" to "ask for the bytes that matter".

COG does it with internal tiles and overviews, Zarr with chunks in separate objects, GeoParquet with row groups and a bounding-box column, and STAC is the catalogue that lets you find the right file before fetching any of it. The trade is real: these formats are written once and read many times, and none of them likes being edited in place.

Below: the concepts cover what each format stores and why the layout matters; the how-to guides read, write and convert; and the fixes are the reads that come back slow, empty or unexpectedly expensive.

Learning paths run through this topic too — see the end of this page.

🧭 Start with the concepts

The ideas the how-to guides below assume you already have.

⚙️ Then do the task

Step-by-step walkthroughs with code you can copy and adapt.

Show every how-to guide in this topic

🔧 When it breaks

The errors this topic produces, with the cause explained rather than just the command.

Related topics

Subjects that keep coming up alongside this one, most overlap first.

Keep browsing

Cloud-native formats sits in Production. Every topic on the site is listed on the topics index.