This message was deleted.

Hi, @clean-businessperson-77391 . As I far as I know many to many rooms are limited by node capacity and the whole room should be located on the same LiveKit server instance. But when we talk about streaming we have one to many relation. Which means we have only one publisher and other participants are able only to receive video. In this case on LiveKit Cloud side they give ability to multiplex publisher’s stream using other LiveKit servers. For example 1 livekit servers, and each is able to serve 1k subscribers, and if we want to stream to 1.5k subscribers, 999 participants can be connected to server 1 and rest 500 will be connected to other server which will connect to first one and will be also sending video. In this case looks like you can scale with no limits from LiveKit server side

@clean-businessperson-77391 the primary difference is: • A LiveKit Open Source session can only be hosted by a single server. • A LiveKit Cloud session can be hosted across many servers. The benchmarks you linked to in docs are for LiveKit Open Source. A single LiveKit Cloud session (i.e. room) can support millions of simultaneous users.

@magnificent-art-43333 Thank you. Is that official information? The cloud version has implemented that rooms scale automatically between servers. But the code in the open source, self-hosted version is limited by the remaining resources on the node where the room was created? Is there a way to connect multiple smaller rooms? Can we manually assign rooms to nodes and connect them using egress/ingress or using webrtc server-to-server? We don't know how big a conference room will be when it's created. Currently we use a solution based on mediasoup where we assign the users to servers and the server can fetch the stream from the origin server (where it is published).

@clean-businessperson-77391 this is official information. You wrote a lot above, but if I’m parsing properly, you’re correct. LiveKit Open Source does not include the relay/mesh functionality. You can build your own version of it (some have), but our version is not open source at this time. There are no short term plans to make it open source or source-available. As of today, this is the only core feature that is unique to LiveKit Cloud. All other core features are free and open source. Cloud also has a few auxiliary features which aren’t available in open source today like real-time analytics/telemetry dashboards and noise cancellation in partnership with Krisp.

Thank you. I understand that this is a sensible decision to keep that mesh capability for yourself. Unfortunately as a volunteer-based non profit organisation, pay per minute and dependency on a proprietary software are not an option.

You can build your own version of it (some have)

Would that mean forking the software or are there APIs that enable building that?

@clean-businessperson-77391 let’s talk more. Introducing you to @full-gold-39399 and @bright-scooter-10854. We’ve long wanted to formalize how we can support non-profit efforts on the Cloud side. Maybe there’s a path there. Fwiw, we don’t charge per-minute like everyone else. We charge based on resource (bandwidth, cpu) usage. Ultimately we need to build a sustainable business, so we can keep working on LiveKit for the decades to come. That’s one reason why mesh is proprietary, but it’s also complicated to operate and would present a large support surface area for us. We continue to automate components of Cloud and our goal is to eventually get to a “fire-and-forget” point. Once we do, we’ll reconsider our stance. There are certainly some interfaces/hooks in Open Source where you can start your own mesh functionality. Our Cloud product is an orchestration layer that wraps LiveKit Open Source, so the hooks are already there for our (and potentially your) usage.

Oh, thanks for being open to support non-profit efforts. We are currently not actively looking for a new solution. I just stumbled upon this project and thought it could be viable alternative as a basis for our system. Our current system is based on mediasoup and runs on physical servers at Hetzner with 10Gbit interfaces. I just checked the total traffic, it showed 75TiB over the last 30d (incl. recording, video-on-demand and server-to-server and backup) Besides that I'm also doing some research for a new project. For that we might also consider liveKit. Your response makes me feel good about supporting you.

Hey @clean-businessperson-77391, great to connect. Just caught up with the conversation and wanted to offer a chat if you're up for it. As Russ mentioned, we're happy to explore options for non-profits with LiveKit. No pressure if you're not actively looking right now – we can share info and answer any questions for future consideration. If you're open for a call this week, let me know a time that suits you.

Hi @clean-businessperson-77391 , just weighing in with my two cents that my team had previously been running our own mediasoup SFU setup with a home-grown orchestrated global network for scaling and we switched to livekit cloud and we are incredibly glad we did. The team is wonderful and the API is fantastic.