Learn

Multicast for AV

Why AV-over-IP needs multicast, how IGMP snooping and a querier work, and the settings that keep streams from flooding the network.

Last updated 2026-07-06

Multicast is the delivery model that makes AV-over-IP scale — and the single most common thing people misconfigure. Get it right and one 4K source feeds fifty displays effortlessly. Get it wrong and that same stream floods every port on the switch, saturating the network and breaking unrelated traffic. This guide explains the mechanism and the settings that tame it.

Unicast vs multicast vs broadcast

  • Unicast — one sender to one receiver. Sending the same 4K stream to 50 displays would mean 50 copies; impossible at AV bitrates.
  • Broadcast — one sender to everyone on the segment. No targeting; floods the network.
  • Multicast — one sender to a group of subscribed receivers. Exactly one copy travels each link, delivered only to ports that asked for it. This is what AV-over-IP uses.

The flooding problem

By default, a switch treats a multicast frame like a broadcast: it floods it out every port. With several high-bitrate AV streams, that instantly overwhelms endpoints and uplinks. The fix is to teach the switch which ports actually want each stream — that’s IGMP snooping’s job.

IGMP snooping

IGMP (Internet Group Management Protocol) is how receivers announce “I want group X.” IGMP snooping lets the switch listen to those messages and build a map of group → ports, so it forwards each multicast stream only to subscribed ports.

  • Enable IGMP snooping on every switch carrying AV traffic.
  • Without it, multicast = broadcast, and the network floods.

The querier (don’t forget this)

IGMP memberships expire unless refreshed. Something must periodically ask “who still wants which groups?” — that’s the IGMP querier.

  • You need exactly one active querier per VLAN/subnet.
  • A router normally provides it; on an isolated AV VLAN with no router, you must designate a switch as the querier.
  • Symptom of a missing querier: streams work for a minute or two, then drop as memberships time out and the switch stops forwarding. Classic “it worked, then died” AV-over-IP fault.

Related settings that bite

  • IGMP version match. Keep switches and endpoints on compatible IGMP versions (commonly v2 or v3); mismatches cause silent join failures.
  • Fast-leave / immediate-leave. Speeds up teardown when a display unsubscribes, freeing bandwidth quickly — useful in matrix-style switching.
  • Unregistered multicast handling. Some switches flood unregistered multicast by default; set it to drop/filter so stray groups don’t leak.
  • One VLAN, one querier. Two active queriers on the same VLAN fight; isolate AV and control it.

Quick configuration checklist

  1. Put AV on a dedicated VLAN.
  2. Enable IGMP snooping on every switch in that VLAN.
  3. Enable one IGMP querier for the VLAN (a switch, if there’s no router).
  4. Match IGMP versions across switches and endpoints.
  5. Set unregistered multicast to filter, not flood.
  6. Verify: start a stream, confirm it reaches subscribers and that non-subscribed ports stay quiet.

Rule of thumb

IGMP snooping stops the flood; the querier keeps it working. If AV-over-IP streams start fine and die after a minute or two, suspect a missing or mis-scoped querier first.

Related

Sources

  • RFC 2236 — Internet Group Management Protocol, Version 2 (IGMPv2).
  • RFC 3376 — Internet Group Management Protocol, Version 3 (IGMPv3).
  • RFC 4541 — considerations for IGMP and MLD snooping switches (snooping and querier behavior).