I've looked at Cisco's VoWLAN Design Guide v4.1 and it has quite a bit of info for Vocera. The controller setup with a multicast address and the WLAN features to properly classify voice traffic is clear to me. I have 'mls qos trust dscp' on all of the AP interfaces. What is missing for me is what setting I place on the trunk ports directly connected to my WLC 4402. Is this where I change to 'mls qos trust cos' or do I leave it 'mls qos trust dscp' and use something like 'auto qos' to create my queues? Or even yet, do I create a policy map and manually map Cisco's Voice DSCP of 6 to 802.1p EF/46?
You mention that all trunks up to the WLC should be set to trust DSCP. Does this include the trunks connected TO the WLC? This is counter-intuitive since DSCP is typically routed/L3 ports and CoS is L2. Since trunks are L2, why would I trust DSCP? For a single controller Vocera deployment, do I still need queues and CoS mappings?
I have to admit that I have seen conflicting documentation between and among Cisco and Vocera regarding this particular setting. It makes more sense to me to apply CoS to those trunks/port-channels as you mention. However, I have tested the configuration in both ways, and it doesn't seem to have any impact on performance in my scenario. The amount of Vocera badges and the fact that everything is directly connected to the 3750G tells me that I'm probably not taxing the system enough to make the configuration differences apparent.