Download
Some OpenPlanetData datasets are hundreds of gigabytes. A standard browser download uses a single HTTP connection, which rarely saturates modern network links. With a single connection, downloading a 175 GB file can take over an hour, even with a fiber connection.
Why Rclone?
Section titled “Why Rclone?”Rclone solves this by splitting the file into many chunks and downloading them in parallel — making full use of your available bandwidth.
Rclone’s multi-threaded mode opens dozens of parallel connections, each fetching a different part of the file simultaneously. This lets you reach speeds that would otherwise be impossible with a single stream.
Here is a real-world example downloading the 175 GB OpenStreetMap GeoParquet dataset over a standard 8 Gbit/s residential ISP connection — this is a remote download, not local:
$ rclone copy \ --buffer-size 0 \ --disable-http2 \ --http-url https://download.openplanetdata.com \ :http:osm/planet/geoparquet/v1/planet-latest.osm.parquet . \ --multi-thread-cutoff 0 \ --multi-thread-streams 64 \ --multi-thread-chunk-size 32M \ --transfers 1 --progress
Transferred: 5.442 GiB / 174.687 GiB, 3%, 763.213 MiB/s, ETA 3m47sTransferred: 0 / 1, 0%Transferred: 6.338 GiB / 174.687 GiB, 4%, 783.920 MiB/s, ETA 3m39sTransferred: 0 / 1, 0%Transferred: 12.964 GiB / 174.687 GiB, 7%, 870.083 MiB/s, ETA 3m10sTransferred: 0 / 1, 0%Transferred: 14.347 GiB / 174.687 GiB, 8%, 876.515 MiB/s, ETA 3m7sTransferred: 0 / 1, 0%Transferred: 15.327 GiB / 174.687 GiB, 9%, 882.108 MiB/s, ETA 3m4sTransferred: 0 / 1, 0%Transferred: 78.810 GiB / 174.687 GiB, 45%, 928.088 MiB/s, ETA 1m45sTransferred: 0 / 1, 0%Elapsed time: 1m28.2sTransferring: * planet-latest.osm.parquet: 45% /174.687Gi, 927.625Mi/s, 1m45sThe file downloads at approximately 900 MiB/s, completing the full 175 GB transfer in around 3 minutes.
Installation
Section titled “Installation”Rclone is available on all major platforms. See the official installation guide or use your package manager:
# macOSbrew install rclone
# Ubuntu / Debiansudo apt install rclone
# Windows (with winget)winget install Rclone.RcloneEvery dataset page on OpenPlanetData includes an Rclone button next to the download link. Clicking it shows a ready-to-use command tailored to that specific file.
The general pattern is:
rclone copy \ --buffer-size 0 \ --disable-http2 \ --http-url https://download.openplanetdata.com \ :http:<remote_path>/<version>/<filename> . \ --multi-thread-cutoff 0 \ --multi-thread-streams 64 \ --multi-thread-chunk-size 32M \ --transfers 1 --progressKey flags
Section titled “Key flags”--multi-thread-streams 64— Number of parallel connections. 64 works well for high-bandwidth links. Reduce to 8–16 on slower connections.--multi-thread-chunk-size 32M— Size of each chunk. 32 MB is a good default.--multi-thread-cutoff 0— Always use multi-threading, even for small files.--disable-http2— Forces HTTP/1.1, which allows true parallel TCP connections instead of multiplexing over a single one.--buffer-size 0— Disables in-memory buffering to reduce RAM usage on large files.--progress— Shows real-time transfer statistics.
Adjusting for your bandwidth
Section titled “Adjusting for your bandwidth”| Link speed | Suggested --multi-thread-streams |
|---|---|
| 100 Mbit/s | 4–8 |
| 1 Gbit/s | 16–32 |
| 8+ Gbit/s | 64 |
Higher values use more CPU and memory. If you see diminishing returns or errors, lower the stream count.