What does your production self-hosted setup actually look like?

Hey guys, seeing a steady stream of “what do I need to run X concurrent sessions?” questions— so let’s build the reference we keep wishing we had.

If you’re running LiveKit in production, drop a reply with any details below…

  • Workload — voice agents, group video, SIP inbound/outbound, or a mix
  • Peak concurrency — rooms, participants at max load
  • Machine specs — vCPU, RAM, single node or cluster, cloud provider and region?
  • What broke first — CPU, bandwidth, TURN, something stranger?
  • Config that delivered the most headroom?

Thanks in advance, I’m still investigating the self-hosting path myself. I’m happy to share my specs when I finish testing.

There is a few load testers around I’ve seen like this: GitHub - thinhdanggroup/livekit-loadtest-tools · GitHub

Just never got around to playing around with it.

Great resource! :clap: I wasn’t aware of these load testers, thanks for sharing the link. Seems like a great place to start.