alt Hacker NewsMy theory is that 80% of workloads on AWS/GCP/Azure are pure waste. They sell complexity-as-a-service. 80% of startups and enterprises could run on a single beefy baremetal server (or two). AWS/GCP/Azure are the result of hype bubbles and VC-funded waste culture, it's not necessary for Europe to recreate that to compete.
Replies
This is an uncomfortable truth on this site, because many of us work for a FAANG company or FAANG partner. If the cloud hadn't grown that much in the last decade or so, the software industry would be relatively unpretentious.
The truth no one wants to hear
Standard Hardware became so powerful over the last 10-15 years, even hightraffic sites from 2010 can be served today with one/two beefy machines.
Thank you. The AWS spaghetti is a trap of unnecessary complexity for most cases. You'd be shocked how far you can scale with a few good baremetal servers running something like Rails and Postgres.
And most (not all) of these workloads are custom software that try to fully reproduce/plumb functionalities that already mostly exist in Unix tools, with worse performance, instead of using/plugging into them.
I once had a working program, running on a 4 GB RAM virtual server with MongoDB. Everything was fast and testing and deploying a new version took me some minutes usually. Existing users were happy as far as I could tell.
But then some corporate IT guy mandated everything had to be using managed AWS services in some three tier dev-test-production setup, despite having no prior experience with that on either side. Cost went up at least 25-fold, the development sucked, new deployments took 30? minutes minimum (because now everything has to run through some build-system I did not control and I had to manually copy keys around every time). I left the company, but I think the product exists to this day with less than 1000 customers. Nothing my 4 GB VS could handle...