[y4mi]

Serverless is at its heart - as I understand it - a dockerized microservice, abstracted away to a degree that the developer no longer has to think about anything but his application code.

You'll definitely be able to ignore it and it probably won't be used in smallish companies for ages.

It's just an easier way to to get your application to scale than homebuild docker images were.

[rhlsthrm]

> You'll definitely be able to ignore it and it probably won't be used in smallish companies for ages.

Why do you say this? I feel like this would be very useful for smallish companies. I'm running eng for my 3 person startup and looking into using Lambda-based microservices with Serverless for our next project. My goal is to completely minimize devops time for our engineers, as well as reduce cost compared to PaaS services.

[xchaotic]

"a degree that the developer no longer has to think about anything but his application code."

a developer still has to understand the implications of resource consumption etc. For performance-critical pieces of code, IMO it's better to have direct access to the hardware - I had a recent first hand experience with this debugging NUMA related performance issue.

[mcintyre1994]

Is there a performance-critical case that isn't already ruled out by the Lambda cold boot problem though?

[y4mi]

nor would you use a dockerized microservice for that, or would you?

serverless is - as i said before - a dockerized microservice as its heart. It should only be used in in places where you'd do it without the abstraction.

There are a lot of services / applications you can build with this. For example Adapters for external SaaS which should be able to trigger certain actions, or just plain JSON APIs which query a DB and output their results...

but using it to join 2 TB of data and process it afterwards in realtime? yeah, thats not a valid usecase for serverless.

[chii]

> no longer has to think about anything but his application code.

i mean, CGI has always existed. This serverless hype is basically rebrand of CGI with some fancy orchestration around autoscaling across boxes (which, tbh, isn't really that much work, and most people don't need the scale required to make this feasible anyway).

[mmt]

> isn't really that much work

I suspect that it's this little parenthetical tidbit, and implicit disagreement with it (or differing definitions of "much") that drives the creation of this kind of abstraction.

In some situations, I consider the gap to be legitimate, where it may be easy (not that much work) for an expert but difficult for everyone else, and, more importantly, becoming expert is non-trivial, even with training/mentorship from one.

In other situations, I consider the gap to be merely one of perception/misestimation, either because it would actually be relatively easy for a non-expert who had actually tried, and/or the needed expertise is shallow enough that it can be quickly taught.

I believe autoscaling is (or at least originally was) of the former category and that the availability of tools and abstractions around it has allowed a broad number of non-experts to leverage the wisdom of a much narrower group of expert practitioners.

OTOH, I believe running hardware in a datacenter (as opposed to outsourcing it to a VPS or even cloud) is of the latter category. I routinely read comments like "have to hire 5 sysadmins" from non-experts when we experts know that estimate is around 20x too high for a scale of hundreds of servers. Even at higher scale, if hiring is necessary, the hardware-specific skills are easily taught, so junior staff is fine.