Can we use Livekit-Server self-hosted with multipl...
# ask-aih
hundreds-yak-62053
12/20/2025, 7:08 AMCan we use Livekit-Server self-hosted with multiple servers for the same room using Redis cache?
t
tall-airline-16756
12/20/2025, 7:08 AMThanks for reaching out, Jasmeet. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.
tall-airline-16756
12/20/2025, 7:09 AMThanks — you're asking whether a self-hosted LiveKit Server cluster can host the same room across multiple servers using Redis.
Short answer: You can run multiple LiveKit Server instances connected via Redis for coordination, but a single room must be hosted on one node at a time. Redis is used for room metadata and inter-node messaging (cluster awareness, stats, signaling), not to split a single room’s media processing across servers.
What you can do instead:
• Run many nodes in a cluster with Redis so rooms are placed on available nodes for load distribution.
• Use region-aware node selection or failover/migration to improve latency and availability; rooms still live on a single node at a time.
If you want, tell me:
• Do you need horizontal scaling of one very large room, or are you aiming to distribute many rooms across nodes?
I can then suggest architecture options (regional nodes, migration, or other patterns) based on that.
Sources: Distributed Setup | LiveKit docs | LiveKit Cloud Architecture | LiveKit docs | How We Built a Globally Distributed WebRTC Mesh Network
h
hundreds-yak-62053
12/20/2025, 7:10 AM• Do you need horizontal scaling of one very large room, or are you aiming to distribute many rooms across nodes?
yes
t
tall-airline-16756
12/20/2025, 7:10 AMThanks — quick clarification: when you answered "yes," did you mean:
1) You need horizontal scaling of a single very large room (splitting participants/media across servers), or 2) You want to distribute many separate rooms across multiple nodes for capacity and load balancing?
Please tell me which one (1 or 2). If it's (1), share approximate participant counts and expected media uplink/downlink (kbps) so I can suggest specific options. If it's (2), tell me expected total concurrent rooms and avg participants per room.
h
hundreds-yak-62053
12/20/2025, 7:10 AM1
t
tall-airline-16756
12/20/2025, 7:11 AMThanks — you're asking about horizontally scaling a single very large room across multiple self-hosted LiveKit nodes.
Clear answer: LiveKit does not split a single room across multiple servers. One room and all its media must be hosted on a single server node at a time. Redis enables cluster coordination (room metadata, signaling, stats, and node discovery) but not sharding of a room’s media processing.
Alternatives you can use:
• Scale up the hosting node (bigger CPU, NIC, and instance types) to handle more participants and higher bitrate.
• Reduce per-participant load with adaptive simulcast/SVC and lower max publish bitrates or reduced resolution for some participants.
• Use selective forwarding / forwarding policies (subscribe filters) so participants only receive needed tracks.
• Region-aware nodes: place rooms on nodes close to majority of participants to reduce latency.
• Implement client-side strategies: breakout rooms, staged attendees (spotlight + viewers), or audience-only viewers to limit heavy upstreams/downstreams.
• Use live migration/failover: move a room to another node if the host node becomes overloaded (room still runs on a single node during migration).
If you want, tell me:
• Expected peak participants and average publish/download kbps per participant, and whether video is required for all participants.
I’ll recommend specific node specs, simulcast/SVC settings, and concrete configuration changes to support your scale.
4 Views