[cottsak]

The challenge with these editorials is that you can never really capture the opportunity cost from the author - they didn't build the monolith without the complex infrastructure as an alternative in parallel.

I wonder how much 'sunk cost' and other psychological factors play into the statement: > We started with Kubernetes a bit too early,

It feels like a "it wasn't that bad". But if you consider the dedicated resources, costs of pain and suffering that may have been avoided with a simpler architecture and infrastructure, I'm forced to wonder if this short comment hides a lot under the covers.

Engineers can have very high pain thresholds sometimes.

[justinclift]

It also sounds like they have several staff just dedicated to managing their Kubernetes cluster, even though they're using a managed Kubernetes service (AKS) for the last few years.

Wonder if they're including those engineers in the cost calculation?

[rwmj]

I also didn't understand what they were building. Is it a simple database-backed website or something truly complicated? How many users do they have? What's the scale of operations? Why wouldn't a handful of dedicated hosts in a datacenter have worked?