I could have 5-8 NUCs/RPis, install a slew of conventional services I was curious about, and consider things like power, cooling, cables, storage.
_or_
I could pay GKE for a k8s cluster, install most of the things I was curious about, and avoid dealing with, e.g., corruption of the SD card. Or the cat knocking the desk-rack of NUCs off.
since rack, stack, and on-prem issues are not what I was wanting to sort out, I chose to deal with what I wanted to learn, rather than rat-hole down data center admin issues.
I've been doing this work for years and years and years. I was scrounging boxes from the CS department for research 15 years ago (or more). Servers are a fundamentally flawed unit of reliability and development. Moving to a design world where you design for someone to knock over the box and _the computation keeps going_ is paradigmatically better.
When you deal with forward looking people in the server-centric world and you show them a service / "keep going" world, their mind is blown and they want it IME.
I will say that in the most part I agree with you. It is better to live in a civilisation, pay tax, have the litter collected, the roads paved, rather than trying to roll your own.
However, being able to, to varying degree, opt out of said civilisation when it runs counter to your ideals or becomes to authoritarian or means that existing within it could compromise what you believe in, is important. The cloud is great, but we must strive to have the skills to be self reliant so it is truly a choice to participate in it to the extent that we do.
I could have 5-8 NUCs/RPis, install a slew of conventional services I was curious about, and consider things like power, cooling, cables, storage.
_or_
I could pay GKE for a k8s cluster, install most of the things I was curious about, and avoid dealing with, e.g., corruption of the SD card. Or the cat knocking the desk-rack of NUCs off.
since rack, stack, and on-prem issues are not what I was wanting to sort out, I chose to deal with what I wanted to learn, rather than rat-hole down data center admin issues.
I've been doing this work for years and years and years. I was scrounging boxes from the CS department for research 15 years ago (or more). Servers are a fundamentally flawed unit of reliability and development. Moving to a design world where you design for someone to knock over the box and _the computation keeps going_ is paradigmatically better.
When you deal with forward looking people in the server-centric world and you show them a service / "keep going" world, their mind is blown and they want it IME.