Post by Yan Cui

This is HUGE! It solves the "Lambda is expensive at scale" problem, and avoids the "having to manage infra yourself" problem with self-hosted FAAS solutions. You keep the same programming model with Lambda, and configure HOW you want AWS to manage your EC2 instances (instead of managing them yourself). With this change, Lambda is able to handle multiple requests in the same execution environment (similar to Vercel's in-function concurrency). Allowing you to make use of idle CPU cycles while your code is waiting on IO. At the same time, it gives you more control around the networking environment, which is also a common enterprise requirement. With this mode, you no longer pay for execution time, but instead, you pay for: 1) Lambda requests - $0.20 per million 2) EC2 cost 3) 15% premium on the EC2 cost (you can use existing EC2 saving plans on the managed instances) As excited as I am about this release (it's one of the big ones I've been waiting for), it's not for everyone. But, for folks who want the simplicity of Lambda, but wants better cost efficiency at scale (e.g. running thousands of RPS), then this is for you! Nice, re:Invent has started with a bang :-D https://lnkd.in/ewwnNaSk