Hacker News new | past | comments | ask | show | jobs | submit login

What's a good way to estimate server requirements?



Sorry I don't have any exact numbers. For fun I threw it on a micro instance and didn't have a problem with 7 people in the call.

I am going to work on writing an Apache Bench for WebRTC this year though. It is really hard to load test/bench this stuff because the tooling just doesn't exist. A lot of people roll out WebRTC, but then go down under real world load.


for webrtc the load is on the clients. the server load is for "poor connectivity" clients. the ones tha need nat transversal and proxing of kinds.

in summary, if you have first class clients you can "host" a million person session in a raspberypi. if you have two bad clients, you better have tons of cpu and bandwidth!




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: