Where this stands
Discovery is done. The plan is about half baked. I’m publishing it anyway because the reasoning is the useful part.
What’s known
- QoS for Dante isn’t standardized end to end across the Nexus and Catalyst switches yet.
- Audinate asked for Dante QoS and for the IGMP querier to be standardized. The two switch platforms weren’t running the querier the same way.
My approach
I pushed back on changing IGMP and QoS at the same time, or right away. Both touch every multicast flow in the plant. Each change gets:
- a written plan
- a change window
- its own validation
- a rollback plan
One change, one result. If something breaks, I want to know which change did it.
Plan direction
- QoS on Dante access ports and the uplink trunks, aimed at protecting PTP timing first. Audio follows the clock.
- Standardize the IGMP querier version and behavior per VLAN as a separate change.
Still to work out
- The exact trust and queuing model on each switch platform. Nexus and Catalyst handle this differently, so the config won’t be a copy-paste between them.
- How to validate the result: what “better” looks like in latency, clock stability, and dropouts.
How I’ll test it before it touches the plant
- Lab first. Try the QoS config on a test path with a packet capture, so I can see the markings and queues actually doing what I think they do.
- Measure before and after. Latency, clock stability, and dropouts on the same devices, so “better” is a number and not a feeling.
- Then one change window at a time, starting with QoS and leaving the IGMP change for later.
Where I am
Discovery came from working through the switches at the CLI. The written plan is in progress. Nothing has changed in production yet, and this page will say when it does.