Stuart Miniman36
What sort of scale can users expect to successfully deploy with #opensstack - most solutions that I hear of are still in less than 100 servers
Scott Carlson
We have a few thousand hypervisors in production across availability zones. it scales just fine :)
cloud foreign
I think @noggin143 might have something to say about scale :)
David Pollack
Further to that point, what are impediments to scaling out to production networks
Sean Winn
I have some customer deployments that run into the hundreds of nodes but @noggin143 can definitely top that
Tim Bell
We're running 75,000 cores at CERN on around 3,500 hypervisors across two datacentres. The nova cells feature is great for this.
Kenneth Hui @rubrikInc HQ
What are some cops challenges at that scale?
Scott Carlson
you have to decide on bridged versus overlay in production. There are performance and availability considerations. also include data center redundancy and availability zones
Shamail Tahir
From the given answers, it scales well. I think some of the success @ scale depends on how you plan up front.
Tim Bell
for us, the biggest challenge has been in the automation. Making sure that procedures are O(1) rather than O(n). Puppet, mcollective and other tools are a great help here.
David Pollack
And probably is unique to each customer type and their objectives
Scott Carlson
@hui_kenneth you have to start thinking at scale from the beginning. consider what you can tolerate in terms of failure from the control plane as well as the data nodes. Do you scale virtual security (fw/lb) or layer-3 physical
Scott Carlson
@noggin143 In addition to scale-up, there are scale-down considerations and people need to get used to the 'destruction' of their environment when you give them a new one
