Kubernetes, the $1M problem??

Over the weekend we saw a twitter post from ChangeLog (below), that reminded us of a number we had heard from multiple other sources too...  we heard "in order to use Kubernetes in Production, you better be ready to add AT LEAST $1M to your IT costs". So seeing this same dollar value represented as "you need to have a million dollar problem" wasn't a surprise to us.

Anyway, when we first heard that "add $1M to your IT costs" comment last year, we dug around to try and understand if it was an off-hand comment, or if there was some logic to it.. turns out there was..

The logic was based around it being a 3 Year TCO, and it was to self-manage a small scale Kubernetes Cluster (scale wasn't mentioned, but for arguments sake lets say under 10 nodes) for a production critical environment (not a non-prod environment, a home lab, or sandpit) that is supported internally under an 11x5 SLA (and on-call out of hours). The logic said that two Kubernetes certified SREs would be needed (either new recruits or existing staff retrained) who would each likely command at least $150k annual salary (clearly this wasnt Silicon Valley numbers). The logic also said that if you use Kubernetes in Production you really should have some sort of support/subscription agreement for it, so there was an allowance of $50k p/a for that (Rancher starts at $60k p/a for a paid subscription, so $50k is actually a bit low).

So, (150k x 2 + 50k) x 3 = $1M.

And that was how the original 1M was calculated.

The original author also justified the number as incremental by saying "if you didn't go with Kubernetes, and stuck with VMs, you definitely do not need this additional cost."

Anyway, we neither agree nor disagree with the post nor the prior commentary, just giving more substance to the claims made by others from what we had heard through our external partners.

Are you running Kubernetes in production? What do you think? Let us know in the comments below.

Neil