April 2016 edition of OpenStack User Survey results are out, with largest participation by far and more insights and analysis than previous editions. Here is my take.
POC to Production Conversion
User surveys show that more POCs are moving to production. 65% of users have reported OpenStack deployments in production, while in previous edition, 59% of users reported so. This reflects increased confidence on OpenStack among customers. This also goes with an increased Net Promoter Score (NPS) than previous surveys, with more users willing to recommend OpenStack. However, I would like to point out the gradual decrease in the % of POCs. Are the new POCs not coming enough?
The % of deployments (both in production and not) using latest releases of OpenStack has increased significantly. Even in the previous user survey, more than 70 % of deployments were using Juno or older versions while the latest survey shows that more than 50%1 were on Kilo or higher. Though all the deployments using latest versions are not version upgrades, we can safely deduct that a significant % of such deployments on newer versions are upgrades from older versions from the decrease in % of older version deployments. This increase in use of latest versions reflects an improved confidence on the OpenStack upgrade process.
Probably the most important data from the survey is the size of OpenStack deployments/ users. More than 50% of deployments have more than 100 users and about 25% of deployments have more than 100 physical nodes. Numbers around cores and virtual machines are more impressive, with more than one-third of the deployments supporting more than 1000 cores or 1000 instances. More than 24% of deployments use more than 100 TB of block storage and about one-third of deployments supported more than 10,000 objects, with few supporting more than a billion objects.
Open Source Alternatives
It was also interesting to see larger adoption of open source alternatives than vendor offerings. For example, OVS and nova-network have more adoption than MidoNet or PLUMgrid neutron drivers or VMware NSX. Similary, Ceph has more adoption than NetApp or EMC or GlusterFS. This demonstrates the success that open source technologies/ eco-system enjoy, with OpenStack playing well with them. This also shows the need for OpenStack to collaborate well with surrounding open source technologies and projects, as rightly called out during keynotes during second day of the Austin Summit.
It is also interesting to note that open source technologies and alternatives have found larger adoption among large scale deployments too. About 34% of deployments with 1000+ cores use OVS and about 45% of them use Ceph, when compared to non open source alternatives.
There were some differing feedback from users on the stability of OpenStack: while some commented that it has improved since Kilo release, some have expressed otherwise. Nevertheless, lack of stability is sighted as one of the reasons why someone would not recommend OpenStack.
Some users have expressed that robust set of features that OpenStack provides bring in fair amount of complexity. This needs to be overcome and could be a detractor
Inconsistency among OpenStack projects has been sighted as another important concern by some of the users. Increasing sprawl of projects, lack of coordination between projects and fragmented governance are resulting in inconsistent levels of maturity and stability among OpenStack services.
I had to list this under Ugly because this has been a constant feedback across many users surveys now. Even though deployment experience has improved a lot and has become more reliable, the fact that this shows up again and again conveys something. This needs to be addressed both through education and improving deployment experience.
Overall, the latest edition of the OpenStack User Survey showed upward swing in adoption and increased user confidence. This is trending in correct direction. I suggest adding following questions to the survey in order to track upgrades and POC => Production conversions and understand the way OpenStack is consumed better.
a. Is this a new POC?
b. Which version of OpenStack are you using? Is this an upgrade from an earlier version of OpenStack? If so, which release?
c. How do you consume OpenStack (DIY, Professional Services, Hosted, or Public cloud?).
d. Which Distribution do you use?
I have used conservative number here, as the %s add up to more than 100%, possibly due to users reporting use of more than one version. I am reaching out to the Foundation to make sure there was no data analysis error here