Post

On the eve of Kubecon I think it's time we acknowledge the cloud hasn't lived up to the hype It is an amazing tool that provides a ton of convenience, but long term the promises turn out to just be marketing hype

I'm not only talking about the cost of the cloud. Although, that turns out to be a big broken promise If you're renting compute from a major cloud provider, your spending significantly more than you may need More details here thenewstack.io/cloud-vs-on-...

All the Kubernetes shifting and "modernization" is also causing long term re-platforming with a lot of added complexity

So many companies are hoping a platform will save them from the complexity without realizing how difficult it is to build a platform and how impossible it is to build useful abstractions on top of a complex and constantly shifting base They're building illusions, not abstractions

Companies have shifted from moving to the cloud to get rid of "undifferentiated heavy lifting" and now they're all doing the same thing - cost optimization - reducing cognitive load

Once you've cost optimized your compute (reserved instances) and reduced your cognitive load (platform) you have a yearly budget spend on servers with a bespoke, lock-in you've built. It's a lot like a data center but 10x the cost

The building blocks are familiar, but the ability to fix your own problems is non-existent because you don't own all the layers The higher level abstraction (eg lambda) the harder it is to hide it

Companies going all in on Kubernetes worry me just the same. They're turning everything into a Kubernetes problem. Maybe that seems like a good idea because debugging will be consistent, but instead it makes things that used to be simple more complex

CAPI makes managing Kubernetes a Kubernetes problem Crossplane makes IaC a Kubernetes problem CRDs and controllers are easy to create and difficult to maintain (kinda like a cloud) The industry depends on them too much and is afraid of doing something different

I've been been impressed with people building things outside of Kubernetes Folks like @systeminit.com and @oxidecomputer.bsky.social not afraid to do something different All of infrastructure isn't going to be Kubernetes or cloud and a lot of people seem to forget

The cloud and Kubernetes are bother going to exist for a long time, but we should stay grounded to what we need I call this “ground computing” The ability to know what you need and build or buy it Less hype, more slow thinking about problems and options

Ground computing doesn’t mean you’re on-prem It means you’re careful about considering what you need You own the lower levels of the stack so you can dig in if you need to

I have learned that "the cloud" is more than just a product or technology for a lot of people For lots of people it's hope It gives them a target to change their lives and get out of bad situations. It helps them learn a widely used technology to be effective at a current or future job