Showing results for 
Search instead for 
Did you mean: 

OpenShift licensing


can someone explain a few points on licensing, I'm finding it very confusing. I've used it before, ran OKD 3.11, am running OpenShift on public cloud (demo account) and my machine (Code Ready). I'd like to suggest it sometimes, but have no idea how it's licensed.

I've read the licensing guide, but found quite a few things unclear:

  • When deploying on private cloud (say RH EV, or RH OpenStack, or whatever) can you only license per core? Is there and option for "per CPU"?

  • Can you, at least, license per CPU for bare metal?

  • Do you pay for Master / Infrastructure nodes, or just for those running you apps ?

  • It says that "hyperthreading is intel-only" - what about AMD? Do I count two cores are one intel core?

  • What about, say IBM Power? If I order a machine from Raptor, does a 8-core 4-way SMT CPU count as just 4 2-core packages?

  • How do you license storage? Per cluster? Per storage node? Storage core? Capacity? Any advantage in doing so, and not just using one underlying on the hypervisor - say Ceph already on OpenStack?

  • How do you license the container registry, if you want to use Quay? One for the entire cluster, or ...?

I've found some docs mentioning per-CPU pricing, but all are dated with OpenShift 3.10/3.11 in mind.

I'm a consultant, and deal with technical issues. I'd really appreciate some insight here - I do need to understand this... :) I like my OpenShift, would like to recommend it, but before I do I need to understand how it's priced.

Labels (1)
1 Reply

Pedro, I'm interested in the same discussion. 

there are two models as of May 2022. On Socket basis (per CPU) for Baremetal deployments, and On Cores basis (per Core) for other Deployments such on vSphere and Clouds.


I am unclear though which Cores are considered, the virtual cores presented to the worker nodes or the Physical Cores that are mapped to the Worker Nodes!  they wouldn't be the same if there are resources overcommitments on the hypervisor level which is common for workloads. e.g. 2:1 or 4:1 

Join the discussion
You must log in to join this conversation.