2019 ROS Metrics Report

We’ve posted the annual ROS Metrics report for 2019. You can download it from here and it’s been added to the Metrics wiki page where you can find links to all the previous versions as well.

Background

We started collecting metrics in 2011. Reviewing the history you can see the growth and evolution of the community.

Measuring open source communities is very hard. The nature of being open and redistributable means that we definitely do not know everyone who is using it and that’s part of being open. These metrics can provide insight into trends within the community but should not be considered exhaustive or even close to complete but as a consistent snapshot. We have public instructions for setting up mirrors and these measurements do not count the any statistics for mirrors either private or public. Public mirrors are listed at Mirrors - ROS Wiki

Updates

Every year we seek to provide the same metrics so that trends can be observed. However we also look to update the metrics to include new statistics or cover new aspects that we think may be interesting trends in the future. This year we have added a new viewing metric which is the viewing statistics for https://index.ros.org . It is part of our revamp of resources to evolve the ROS infrastructure as we continue to need to scale upward.

Commentary

At a high level the growth trend continues for the ROS community.
We are now regularly providing around 8,000 GB of package downloads each month.
Last year we provided over 72,000 GB of downloads, and we’ve almost reached that point already with 67,000 GB already download at the beginning of October.

One trend that has been continuing over the last few years is that the growth of ROS in China has continued to the point that they are now the largest userbase by country.

There are two metrics in this report that appear to be different than the general trend.
The first is that wiki edits are significantly up from last year. I suspect that this might just be a sampling artifact. If a few people are particularly active for a few days, especially if there’s a refactoring of a wiki page or two this can quite easily update the numbers.

The other number that has grown a lot is the answers.ros.org users number. The site is currently under a spam attack and through the generous support of our many moderators we have successfully prevented the spammers from posting to the site. However they have broken every captcha effort that we’ve been able to throw at them and so the spammers continue to register but remain with their accounts in a probationary status. When looking at these numbers I suggest considering the users who have contributed to the site and have non-zero karma.

Related Work

I’d also like to highlight again the work of @DLu who has also put together a site for viewing longer term trends of various ROS metrics. It’s a great complement to this annual report. Please see his announcement

And visit the site at: metrics.ros.org

6 Likes

Great work, interesting to see. Are there any metrics about operating system usage trends?

The linux distro usage is here: https://metrics.ros.org/packages_linux.html

Tracking non-linux OS has been fairly noisy on the metrics I’ve looked at, but hopefully there will be clear trends with ROS2 someday.

@tfoote Is there a place other than packages.ros.org where binaries are downloaded from for things like Windows?

The Windows and OSX binaries are hosted on GitHub under the ros2/ros2/releases

The download metrics are available via the GitHub API. Here’s an example of scraping them.

1 Like

Thanks @tfoote. Interesting – a quick glance through that linked site appears to show Windows is slightly more popular than MacOS for ROS2…Didn’t expect that.

Integrated with metrics.ros.org: https://metrics.ros.org/packages_ros2.html

1 Like

I wonder if there’s some sampling bias here if there are a few large build farms / cloud services running Windows downloading packages. I can’t possibly believe that Windows is actually used more than Linux in most of these distributions. I’m not sure who would deploy robot fleets running windows (not a down on Window’s for any particular reason more than Linux is free and deploying assets at scale that $100+ is substantial on BOM) and given the heritage of Linux in ROS, it seems suspect to me that people moving from ROS1 to ROS2 would immediately jump back to Windows-land after years of working in Linux.

My comment was Windows vs OSX, not vs Linux. Linux still dominates. That said, it was a very quick glance through a couple recent releases, not an exhaustive study of the trends across all releases.

EDIT: Also, comparing against Linux is a bad comparison here because I’m sure most Linux downloads are through apt, not github releases. I figured comparing Windows to OSX was reasonable since neither have strong package management, and users would be more likely to grab their binaries from github on these platforms.