What's OpenStack's killer app? Users say it's dev and test.
According to the latest OpenStack User Survey, most deployments of the open source cloud infrastructure project are on-premises private clouds for dev-and-test work that serve teams of fewer than 100 users.
The survey tallied responses from users running some 260 deployments of OpenStack worldwide, with results available through a portal that allows the data to be tabulated in various ways.
A companion report released by the OpenStack User Committee distilled the most striking results, but poking through the data in the portal provides a more granular look at what OpenStack users are doing.
Build it in house, use it to test
The User Survey report prominently cited the growing number of OpenStack deployments that are considered "production" or "full operational use" -- 71 percent as of September 2016, up from 59 percent the same time last year and 32 percent in late 2013. Likewise, the number of installations marked "proof of concept" or "in testing" have shrunk from 2013 highs of 34 percent for both down to 11 percent and 18 percent, respectively.
Current OpenStack users also are less likely to badmouth the product to others than they were last year. The number of users that are "detractors" has dropped from 21 percent in 2015 to 10 percent currently. But those users, by and large, haven't become "promoters," which is around 53 percent of the surveyed userbase; instead, they're joining the ranks of the "passives," who have a neutral recommendation stance about OpenStack.
The data also clearly show where OpenStack deployments are going and to what end. Sixty-five percent are used to create on-premises private clouds, which are in turn being used mainly for dev-and-test work (65 percent) and "infrastructure services" (51 percent). The third most common use case, databases, comes in at 35 percent, with all other applications -- web services, network function virtualization, analytics -- all trailing in percentages after that.
A little more than half of the deployments out there service fewer than 100 users, with 20 percent serving one to nine users, and 37 percent servicing 10 to 99 users. The next largest slice, 27 percent, was for teams of 100 to 999 users, but everything beyond that is 11 percent of the pie or less.
The Kubernetification of OpenStack continues
Of all the recent changes made to OpenStack, the biggest is how it's been incrementally reworked around the popular container orchestration framework Kubernetes. It's still too early to tell how this will continue to steer OpenStack development and use, but a big slice of OpenStack users have been primed since the last survey to make good use of Kubernetes. As of right now, some 48 percent already use it for PaaS in OpenStack.
OpenStack users are also big fans of leveraging other common enterprise infrastructure tools over those found in OpenStack itself. Ansible and Puppet (43 percent and 40 percent) were the two most common tools for cluster deployment, with OpenStack's own Fuel at a mere 17 percent.
OpenStack had made its most noticeable inroads in telecommunications companies because its networking functionality has proven it can provide a viable alternative to the costly proprietary hardware often used in telecom applications. Telecom is still the second-largest industry sector identified by OpenStack users (13 percent), although general "information technology" companies make up some 68 percent of the total user base.
That said, previous versions of this report have included many companies as part of the 68 percent with telecomlike duties -- "cable TV and ISP," "telco and networking," "data center/co-location" -- so it's tougher to tease out at a glance which of those are "pure" IT shops. (The current version of the report doesn't provide such details.)
When originally unveiled, OpenStack was touted as an alternative to proprietary cloud environments like AWS or VMware. If its most suitable use case (aside from telco infrastructure) is automating dev-and-test environments, the next place we'll likely see confirmation of that is not via self-reported use cases, but through OpenStack's feature set evolving to satisfy those needs.