Scaling Decentralized Low Latency Infrastructure with Huddle01
summary
Huddle01 describes a modular dRTC network that uses centralized SFU-style media nodes plus a registry and reputation-based QoS to keep round trips under 100 ms, avoid hairpinning, and minimize public hops. Implementation uses go-ds CRDT (Badger + IPFS light + libp2p gossipsub), an orchestrator called Sushi orchestrator, and planned user-hosted media nodes. The system runs in production for 2 years with 4 million user-minutes and 75,000 meetings, and the design claims scalable low-latency operation.
AI-written summary · about
chapters (10)
- 0:00Introduction to low latency RTC
- 1:11WebRTC protocols explained
- 4:50WebRTC parallels with gaming
- 6:39Scaling issues with P2P WebRTC
- 7:44Connectivity challenges in WebRTC
- 10:55Media node architecture
- 12:21Quality of service and registry
- 14:06Orchestration for flexibility
- 16:11WebRTC in multiplayer gaming
- 16:54Huddle01 infrastructure results
related talks
08mDePIN: Pushing Decentralization Beyond Blockchain
18mRTC3
30mJitsi Videobridge: the state of the art SFU that powers Jitsi Meet
23mIntroduction to Media Streaming Mesh
25mDisaggregating Persistent Memory and Controlling Them Remotely
25mLIVE Q&A #1
15mReverie: Low Pass Filter-Based Switch Buffer Sharing for Datacenters with RDMA and TCP...
24mScalable Small Message Aggregation on Modern Interconnects