About a year ago, Stephen O’Grady of Redmonk published a comparison of the community metrics of the major configuration management tools. It’s a good read, and I won’t rehash its points. Go read it first.
Today I’d like to take a look at where Ansible is, a year later, using last year’s report as a benchmark. I think it’s fair to say that we’ve done pretty well for our users in 2014.
Debian’s Popularity Contest is an opt-in way for Debian users to share information about the software they’re running on their systems. Although it represents only a small sample of the Linux distro world, it’s useful because it’s one of the few places where we can really see an apples-to-apples comparison of install bases of the various tools.
First, though, a note about the original Popcon analysis: for Ansible, it was an apples-to-oranges analysis. Stephen’s report compared the Ansible control tool — the functional equivalent of a “server” — to Puppet/Chef/Salt agents. When comparing the Ansible package to the server packages of the other configuration management tools, the picture is quite different, and more in line with what we’re seeing elsewhere:
Strong growth demonstrated by Ansible in 2014. Important caveats about the above chart:
- The Puppet line shows far more variability than do the other lines, which may be an artifact of collection method.
- It appears that Chef has moved away from the distro distribution model for their server, so they are significantly under-represented. libchef-ruby was chosen because it appeared to be the best proxy. (If anyone has a better “chef server” package for Popcon comparison purposes, let me know.)
- Ansible is also under-represented to some degree, since a significant amount of Ansible’s user base installs Ansible from PyPi.
- Puppet and Salt could also be under-represented for similar reasons.
Because of Ansible’s agentless model, it’s impossible to get a realistic picture of how many systems are under active management by Ansible. However, If we look at “systems immediately addressable by each configuration management system”, then we would use openssh-server as our “agent”, and the comparison looks like this:
Obviously, not every system that has openssh installed is currently being managed by Ansible. Still, I include the graph because it’s a compelling picture of the power of not having to bootstrap an agent.
Stephen tracked three metrics last year for the four projects: stars, forks, and merged PRs in the last 30 days. A key caveat for all of these graphs: because Ansible and Salt are both Github-native projects, the Github numbers for both projects are understandably higher than their older counterparts.
First, stars. Ansible has opened up its lead from last year, and now has 2-3x more stars than its counterparts.
Second, merged pull requests in last 30 days. This graph looks very similar proportionally to last year, and represents different stages of maturity and different upstream philosophies among the projects. In number of lifetime contributors, Ansible (935) and Saltstack (934) significantly outpace Chef (363) and Puppet (344).
Third, forks. Ansible had a slight lead at the end of 2013; that gap has widened in 2014.
Hacker News Jobs
In his analysis from last year, Stephen referenced a metric he called “Hacker mentions adjusted for growth.” I was unable to replicate that metric, so instead I used Ryan Williams’ excellent “Hacker News Hiring Trends” metric, which pulls from the “whoishiring” discussion threads. It’s a narrow metric, but it shows steadily growing demand for Ansible knowledge among the Hacker News crowd.
Caveat: neither “Salt” nor “Saltstack” appear to be tracked in Ryan’s dataset.
Indeed.com Job Trends
On last year’s chart from indeed.com, Ansible was practically invisible. This year’s chart shows significant growth, even though there’s a long way to go to catch up with incumbents Chef and Puppet.
Caveat: Stephen’s original search used “technology” as the term to try to weed out extraneous data. In my search I used the term “devops” instead, which shows similar results but also allows for the inclusion of Salt.
LinkedIn.com User Groups
Again, as a function of time, Ansible lags behind the incumbents Chef and Puppet, but has made significant strides in the past year, almost tripling the number of users in Linkedin.com user groups.
The total number of StackOverflow questions tagged with the different terms. Interestingly, although Ansible’s numbers are higher here than last year, there have still been fewer questions asked about Ansible than about any of its competitors.
Given the evidently increasing popularity of Ansible along all other metrics, it’s a curious stat — but I like to think that the gap can be explained by Ansible’s ease of use, strong documentation, and large and helpful user community. That explanation may even be reasonable.
Ansible demonstrated strong growth in 2014, and people have noticed. Thoughtworks moved us rapidly from “trial” to “adopt” in 2014 in their technology radar, with some very kind words of endorsement. Opensource.com ranked us as one of the top ten open source projects of 2014.
Which means that the bar has been raised. SDTimes called us the #1 company to watch in 2015. We’ll see about that.
As always, our success is a direct reflection of the success of our passionate community of users and contributors. Thanks to all of you. We’re looking forward to a great new year.