Lessons from the OpenStack user survey

No readers like this yet.
A person holding on to clouds that look like balloons


Every six months, the OpenStack Foundation reports on the results from its user survey. The results for the most recent iteration were released earlier this month on the OpenStack Superuser blog. Let's take a look and see what's new.


First, it's clear that there have been some significant changes to the survey itself since the last set of results were released at the Atlanta summit. The methodology has changed, as has the way the results are presented, making side-to-side comparisons between this survey and the last one just a little bit difficult.

For example, for the past two years, the survey has been primarily administered and analyzed by the OpenStack User Committee along with help from foundation staff. Because of the time commitment this required, this time around a private third-party firm was engaged to help with the survey results.

In addition, a number of the questions changed slightly, and the analysis were only conducted on recent respondants' answers, whereas in the past some cumulative data were used. 

Key findings

So what's new? First, it looks like the number of OpenStack clouds surveyed in active production are up from previous surveys, as cloud administrators moved on from the Dev/QA or Proof of Concept phase into something actually being used by their teams. A plurality of production clouds, and a majority of clouds not-yet-in-production, were deployed on-premise, with others choosing hosted private clouds, public clouds, and hybrid clouds.

Not surprisingly, two thirds of OpenStack deployments were being used by businesses in the IT industry, with telecommunications and academic/research uses each making up about a tenth of what's left, and all others representing around 15%.

While the United States is the primary location of OpenStack clouds, it is not an outright majority, and OpenStack is in use on six continents. Most clouds were running the Havana release, at 45%, while the newer Icehouse represented 37%, and a collection of older releases making up most of what's left, with about 9% of users running a continuous deployment of the latest trunk.

A number of deployment configurations were surveyed. In terms of hypervisor, KVM, QEMU, and ESX were the top three in use. For storage, Ceph, LVM, and GlusterFS were the top three. For deployment, Puppet, Chef, and Ansible topped the list. And on the operating system side, Ubuntu, CentOS, and Red Hat Enterprise Linux topped the list.

Of course, more detail is available in the full results.


So all of this is interesting, but what does it tell us? Like any good study, it would be impossible to end without saying "more research needed."

The user survey gives a snapshot of what's going on in the OpenStack universe, and it might help us discover some trends which can help developers and cloud providers to make better decisions about where to focus resources to meet the needs of cloud users.

Does it answer questions about "which project is most important?" or "who is winning the cloud?" Well, not really. Crosstabs matter, and for a number of reasons, including the privacy of the respondants, this information isn't currently available. We don't know, for example, whether all of the large production systems are running one operating system while the smaller users are running another. We don't know, again for example, whether the decision to use a particular storage driver had more to do with size of deployment or type of application using the storage.

The cross-tabulations might tell us more. So here's to hoping that enough cloud deployments in the future will take the survey so that additional information can be released without compromising any one provider's privacy.

Do you operate a cloud? Want to take the survey? It's always open, and available online on the OpenStack Foundation website.

User profile image.
Jason was an Opensource.com staff member and Red Hatter from 2013 to 2022. This profile contains his work-related articles from that time. Other contributions can be found on his personal account.

Comments are closed.

Creative Commons LicenseThis work is licensed under a Creative Commons Attribution-Share Alike 4.0 International License.