from Hacker News

Currently our stuff runs on two t2.small servers. One serves the app (has everything). I know we can break it up. What we send out is an app which is just a static build (react) and it can be smaller (it's big right now, need to purge Tailwind CSS). We use a postgres backend with Express.

Aside from that the video is handled by another party and our second server is primarily for an Apple CUPS instance. This server also handles websocket connections between users.

The servers are single CPU, 2GB RAM, not a lot of storage (not really needed for operation, we use S3 to store things).

Some ideas to scale are to just serve the front end builds via a CDN/from S3.

I just was trying to figure out how I could get numbers per server, how many people it could hold. I have seen crazy figures like a single NGINX instance doing 10,000 requests out of the box easily.

The printer is the interesting one because the load is in bursts. Not sure what I should cap each machine at even though it's probably never really 100% used at any time.

I've also added full e2e testing with selenium so I can watch it execute a usage and see the staging servers respond in real time but idk how accurate that is/can I just multiply/divide out the resources.