[Openstack-operators] Openstack team size vs's deployment size

Jonathan D. Proulx jon at csail.mit.edu
Mon Sep 12 17:00:35 UTC 2016

I agree this would make a very interesting OPs session.

As many have poitned out it's difficult to really quantify in a
comparable way given the range of roles and approches.

We have 1 region (going on 2) with 100 hypervisors providing a basic
IaaS service (agian looking to expand into various XaaS overlays).
Typically running 600-900 VMs in support of several hundred users &
some randomly shifting number of projects (164 had instances run in
the past month, though some were only single test instances).

We have zero dedicate staff for OpenStack so coming upt with FTE
numbers is guess work at best.

For front end user support, answering questions and using the admin
interface for adds and changes we probably spend about 0.5 FTE (Across
a team of 6-7 people)

For backend stuff for both OpenStack and Ceph probably 1 FTE, maybe a
bit more, mostly handled by 2 individuals also included in the count

This is workable in ideal conditions but far too thin and fragile in
real conditions.  One of the two core OpenStack people moved and it
took 6mo to get a replacemnet. This made us loose an Upgrade cycle as
there was insufficient staffing to test and implement an upgrate to
Liberty during the gap.  This along with otehr bits that didn't happen
accumulated a lot of technical debt in a short time.  We are digging
out but 1 year later we'll be back at current software version but
with zero progress on previously planned service expantions.


More information about the OpenStack-operators mailing list