> IMO what makes something serverless is that you don't need to provision or manage infrastructure, and also no need to worry about scaling as it will go up and down as needed.
This is why serverless is such a stupid term to begin with, IMO. It's not even an accurate descriptor (the stuff is running on a server _somewhere_), so it really could mean anything.
Something like "cloud functions" to describe Lambda and "managed Kubernetes" for Fargate. Of course, these terms are less sexy, so they'd never catch on.
With that definition, wouldn’t something like beanstalk qualify? I think that definition is overly broad. However, you’re right in that it does make me wonder where the definition should start or end.
Re:The Serverless Revolution Has Stalled
by different_sort on Monday October 12, 2020 @08:56PM
(#24761243)
You're responsible for managing the server in beanstalk, but it's not far off. I suppose it's just an earlier generation of the same idea. WHat I mean is, beanstalk runs on EC2 instances that you are responsible to make sure are patched, configured correctly and have everything you need for your application to run.
Fargate/Cloud Run are probably better serverless examples - you bring a container and config and the cloud provider will handle the rest.
I've never used AWS Beanstalk but if it's like Heroku I don't think it qualifies as you are still provisioning infrastructure (much like with EC2).
It provisions EC2 for you, and scaled up or down. To really simply, it’s the basic equivalent to having an infrastructure admin turning EC2 instances on and off as load goes up or down.
IMO what makes something serverless is that you don't need to provision or manage infrastructure, and also no need to worry about scaling as it will go up and down as needed.