As an AWS focused solutions/systems architect, I’ve been feeling this for the last 10ish months too. I attended the first 9 re:Invent conferences (up until Covid upended things) but I was glad I didn’t attend last year; and re:Inforce sounds like it was even worse.
But in the mean time EC2 hasn’t gotten a meaningful feature that’s not to accelerate training or inference since GP3, and people folks are backing away from serverless-first designs because cost-control and other features we’ve been screaming for for several years aren’t being addressed.
Edit to add: on the EC2 side I forgot we got Gravaton3 processors like 18 months ago. That was appreciated for sure.
Aws serverless is a scam also. Lambda with step functions just to run some functions in order without having to scale servers?
Its just ridiculous how much complexity it adds, and the performance is abysmal compared to local function calls as well. It’s like younger devs have never written a fast application without aws.
In theory, running a serverless function can provide adequate response times at costs that are unreachable with private servers. It’s basically those services that would run your application for few minutes every time it received a request, but with theoretically lower overhead since it’s supposed to be a function instead of a full application.