Hacker News new | ask | show | jobs
by Jormundir 301 days ago
-
3 comments

> Building a scalable systems doesn't take extra time anymore,

Except you bleed money all over the place to do it this way.

AWS (and every other cloud provider) is still pricing its product like its 2013

> you just need an engineer who has experience doing it to lead the project

And then you become Figma, who probably should have gotten off of AWS before or with their IPO, rather than doubling down on AWS costs (something like 300k a day).

Most orgs who get past "survival" cant tell me the cost per user, or cost per customer or what each (named) customer is costing them in their stack. If they put in the effort to figure this out what they tend to find is that they are hemorrhaging money to third parties.

I know plenty of CTO's, vp's and Team leads and SR engineers who dont know what the rent vs buy calculation is, never mind how to do it. But I do know that AWS is what props up amazon and its value.

Scale in this case isn't just about software scalability though, in the sense of RPS it can handle. It's:

- Don't build some automated support system until you get enough customers

- Don't build a fully automated provisioning system until it becomes a problem

- Don't build some fancy multi-region failover setup

- Don't worry about designing for some future feature that may or may not be needed

This doesn't mean write bad code though.

-
Cloud is only a very limited and literal part of scalability