Open Stack

Hi folks,
Some (/almost all?) of you have received email inviting to update the survey.

I have went through user survey results of April 2016 [1] and have few comments/suggestions.

1.  "What tools are used to deploy/configure OpenStack clusters?"
I think here there is confusion between 'higher-level deployers' and 'deployment tools' (Ansible, Puppet, Cobbler)
Yeah, Ansible can be used without higher-level deployer [but then OpenStack Ansible could be considered, probably Puppet. 
But in case 'higher-level deployers' and 'deployment tools'  - then once you use one higher-end end deployer you are not able to use another one
 (=you will not use Fuel+RDO)
-So Fuel and Packstack is using Puppet. In general seeing Puppet as #1 with 30% share in production - I am not sure what is really used. 
I suspect that it is actually one  of 'higher-level deployers' using Puppet that is used by many of those users
-I think that RedHat (RH OSP Director+RDO manager as one product?) and maybe HP HLM should be there as high-level deployers?
 
So one of approaches would be to split questions into 2 and actually in case someone selects Fuel (or) it might be OK not to answer about 'deployment tool'

2. Question  "Which OpenStack Compute (Nova) hypervisors are in use?", page 39
Here the replies are communicating that KVM is used in 60% of production environments and QEMU in 10%
My rough estimation is that:
-99.9% of those who use KVM are using QEMU as well (because otherwise it requires some other SW)

And then using QEMU without KVM will have quite a virtualization penalty and I suspect that 10% in production is not the reality 
What I suspect is that some of those who are actually using KVM+QEMU are selecting 'QEMU' or selecting 'QEMU+KVM' 
while in current wording only 'KVM' should be selected by those users according to 'expectations of the survey'
 So I think the option in survey should be 'KVM and QEMU' and 'QEMU (without KVM)'

3. Question "Which OpenStack Network (Neutron) drivers are in use?"
So here confusion comes that on particular OpenStack deployment  [in case Neutron used] will use both ML2 plugin is used as well as one more mechanism drivers.
The current option lists both ML2 and things that are mechanism driver.  
That's a bit fishy that OVS takes like 40%, while I think that like 90% of those are actually using ML2 (therefore 27% is underestimate).
This is in line with [2], but I think that options should be 'ML2 plugin with OVS mechanism driver', 'ML2 plugin with Linuxbridge mechanism driver'
(etc., etc. for other mechanism drivers)
And then to have 'ML2 with other mechanism driver' but never 'ML2 plugin' itself
And SR-IOV is not covered (that might be the place to get data on share of those using SR-IOV)

4. Question "Among clouds that support compatibility APIs, which APIs are supported?
Maybe the question will be 'which compatibility APIs are _used_'? I think that many OpenStack deployments are supporting EC2 compatibility API just because default nova.conf is (/was?) like this , but are not using that. 
Otherwise it will be quite-bad news to understand that EC2 is on deprecation path in Nova.
5. "Which databases are used for OpenStack components?"
I think it is not straightforward to understand whether MySQL is used with Galera vs. MariaDB vs. other MySQL 'flavors'. I think users could be confused (/CLI command can be provided for someone to check on their system?)
6. " What's the size of this cloud?"
I think it is tricky for developer/SW test engineer to answer 'what's the size of the cloud'. In case person works on development of eg. Fuel this can range eg. from 2 to 200 Compute Hosts, but that is not clear how
to put that in survey

There is something so to say in common in most of those possible ambiguities above.
That is the users filling out the survey might not know the correct answer to the question about their deployment(s). 
Competence of users should not be blamed for that. I think that there should be hyperlinks ('how can I know') - indicating typically CLI commands to get that information.
Maybe there could be automation on that based on some existing ('execute script and get information about your cloud')


BR, 
Konstantin
[1] https://www.openstack.org/assets/survey/April-2016-User-Survey-Report.pdf
[2] http://lists.openstack.org/pipermail/user-committee/2016-June/001055.html