Each year, members of the Ceph User Survey Working Group build a survey of Ceph users and conduct the survey in the first calendar quarter of the following year.
The group chooses the questions which tend to be oriented toward better understanding how people use Ceph, what they like about it, what needs improving, and what new features ought to be prioritized. The survey, the conducting of the survey, and the analysis of the results are all done in ways that are aligned to the ideals of the open source community. You can find the survey and results on the Ceph website along with past results and analysis.
We don’t cover the full survey in detail here but instead provide some interesting highlights and analysis below.
We 👏 love 👏 Ceph
Ceph users are happier with the software. The Net Promoter Score ticked up again, another two points on Ceph overall between the previous year and this one, to 49. If you make some assumptions about which users are paying for a supported Ceph deployment from one of the vendors represented, that number goes up to 67, a 37% difference in satisfaction between free users and paying ones. Next year, maybe we should ask more explicitly about those who pay and don’t so we can make fewer assumptions. From a product perspective, it would be nice to assert the difference in experience and outcomes for users who have a relationship with Ceph experts. Data-wise, though, I understand that it could also be explained away as confirmation bias for the purchase decision. Next year, we might dig deeper into the data to tease that out.
Nearly everyone is using Ceph for block storage
When we ask whether you use a particular storage service, and then we count up the responses of yes vs. no, we see a significant positive change for RADOS Block Device. I speculate that this is related to the increase in usage of Openstack and Kubernetes platforms that work well with Ceph’s RBD.
An oddity between 2021 and 2022 is the change in respondents using Ceph to provide multiple kinds of storage services. 10% more users reported delivering two storage services with their Ceph clusters in 2022 than 2021, but 5% fewer reported delivering all three. Could it be related to the decline of CephFS and increase over the years in RGW and market trends toward global collaboration systems that tend to work nicely across hybrid storage implementations and preferring to use S3?
Faster uptake for Pacific vs. Octopus
When we took the survey last year, 48% had already adopted Octopus to some extent. We tend to take the survey around the same time each year, and we have had consistently timed major releases (more or less) in the past. This year 60% had already adopted some form of Pacific. It’s important to note that many of those in both cases were running multiple versions of Ceph.
Where did everybody go?
We had 127 responses to the user survey in 2022, which is a nearly 50% drop in responses year over year. We looked at changes in respondent demographics year over year across things like cluster size, geography, version of Ceph in use, and others, and we did not find significant swings that would indicate a particular population group that just didn’t respond this year.
A few hypotheses...
Hypothesis A – it was too long, and that’s what stopped a bunch of people from tuning in.
Hypothesis B – you didn’t take the survey because you didn’t hear about it.
Hypothesis C – low motivation. It’s possible that people didn’t respond because they don’t see the benefits of their perspectives reflected in the priorities of the community. That may be more a problem of communication than reality though.
Here are just a few examples from Josh Durgin of the way your feedback gets reflected in real action:
Driven by community interaction, including data from the user survey, the Ceph engineering team has prioritized development process improvements. Increased focus on reliability at scale included larger and more realistic scale testing for the Quincy release. Release candidates were published for users to try out and report issues before declaring a stable release.
The survey influenced Red Hat hiring a technical writer to focus on enhancing documentation both in current and prior stable releases.
Zac Dover and Anthony D’Atri, again influenced by past surveys, have both made improvements to Ceph documentation by themselves and through encouraging contributions by others.
Note: I originally co-wrote this post for a different blog with my colleague Craig, but as it is no longer available there, I'm re-publishing it here for posterity in my role as a Ceph Ambassador.
Comments