AV-over-IP: Multicast Freezes After Adding 4K
An AV-over-IP distribution system on a dedicated 1 GbE network has 24 decoders and 8 encoders. Each 1080p stream is about 20 Mbps; six new 4K encoders are added at about 60 Mbps each. The core is a managed Layer 3 switch; two access switches are connected by 1 GbE uplinks. One small unmanaged switch serves a conference room with three decoders. IGMP snooping is enabled on the managed switches, but no IGMP querier is configured. After the 4K sources go live, several decoders show black frames or freeze, especially in the conference room; the problem worsens when more decoders join the same 4K stream. Link utilization on the core uplink peaks at about 45%. The vendor says to enable QoS and buy 10 GbE. You need to determine whether bandwidth is the real limit and what to fix first.
1. Is bandwidth the real bottleneck? Support your answer with the given numbers and the behavior described.
Read the worked answer
The raw unicast-equivalent load is 8 × 20 Mbps + 6 × 60 Mbps = 160 + 360 = 520 Mbps. That is more than half of a 1 GbE link, but the measured core uplink peaks at about 45%, or roughly 450 Mbps. If every stream were replicated to every decoder, the load would be far higher and utilization would be near saturation. The fact that utilization is only 45% while decoders freeze argues against a simple aggregate bandwidth limit.
Multicast changes the calculation. A single multicast stream should be replicated only to ports that requested it; with correct snooping, ten decoders watching one 60 Mbps 4K stream do not consume 600 Mbps on the uplink, but they do consume 60 Mbps on each access link that carries the stream. The conference room symptoms and the worsening when more decoders join the same stream point to multicast flooding or failed group membership, not a need for 10 GbE. QoS may help protect control traffic, but it will not fix a stream that is being flooded to every port.
2. What is the likely multicast configuration fault, and what would you change?
Read the worked answer
IGMP snooping needs a querier to learn which ports belong to which multicast group. With snooping enabled but no querier on the VLAN, the switches may not maintain group membership correctly and can fall back to flooding multicast to all ports. The unmanaged switch in the conference room is a second fault: it does not understand IGMP and will flood multicast to all three decoders and back toward the uplink, wasting bandwidth and causing the reported freezes.
Fix in this order:
- Configure an IGMP querier on the managed network, typically on the core switch or a dedicated router, for the AV VLAN.
- Replace the unmanaged conference-room switch with a managed switch that supports IGMP snooping and fast leave.
- Verify that encoders are sending to the intended multicast groups and that decoders join the correct groups, with no duplicate group assignments.
- Enable IGMPv3 if the platform supports source-specific multicast, and configure fast leave so a decoder leaving a group does not keep receiving traffic.
- Only then apply QoS to prioritize control and audio, not as a substitute for correct multicast.
This addresses the symptom pattern without buying unnecessary 10 GbE.
3. How would you validate the fix and confirm that the network can support future growth?
Read the worked answer
Baseline the network before and after changes. On the core and access switches, record multicast group membership, port utilization, and flood counters while decoders join and leave one 4K stream at a time. Confirm that the stream appears only on ports with active members and that the conference room switch forwards it only to the requesting decoder. Use a packet capture at a non-member port to prove that flooding has stopped.
Then run a realistic load test: all 1080p streams active, all six 4K streams active, and multiple decoders joining the same 4K group simultaneously. Watch for packet loss, latency, and freeze events, and check that the uplink remains below a planned threshold, not merely below 100%. Confirm that fast leave releases bandwidth promptly when a decoder changes channel. For future growth, calculate worst-case per-link replication, not just the sum of stream bitrates, and identify which uplinks would need link aggregation or 10 GbE. Document the querier, VLAN, IGMP version, and switch models so the design is maintainable.