I was forwarded this as gospel, but find it hard to swallow. There may be "some" truth to it, but technically different than this. Anyone know for sure ??
Cisco Voice VLANs...Why Spanning Tree is a Killer!
Ouch! Poor voice or video quality...dropped calls and sessions, etc....the enemies of Voice over IP, IP Telephony and other real-time applications in Cisco Multi-layer Switched Network Environments.
Did you know that the VLAN 'number' that you assign to the Voice VLAN and the number of VLANs that exist on the switch have everything to do with the speed by which PVST or Rapid-PVST will converge your network and allow traffic to flow after a failure occurs? Even in a fully redundant network, with all timers set properly, EIGRP operating perfectly, HSRP implemented with sub-second failover and Spanning Tree as optimized as it can possibly get, if your Voice VLAN numbers are higher than other VLANs on your switch, they will converge last....and you will feel the impact immediately. Even with 802.1w (Rapid Spanning Tree), Cisco implements a 100msec throttle delay on each VLAN when communicating information about VLAN loss to the routing process (EIGRP/OSPF). Therefore, if you have 10 VLANs on an access switch, upon failure of an uplink, fiber traffic on the tenth VLAN will converge 900msec after the first.....Ouch! 900msec of downtime on a VoIP network can be a catastrophe.
So how do we keep the convergence time from affecting our real-time applications? Simple. To ensure optimal convergence for voice or other real-time applications, VLAN number assignments should be mapped such that loss-sensitive applications, such as voice, are assigned the lowest VLAN numbers on the switch or physical interface. This will allow the Voice VLANs to be converged first, thus eliminating the effect of the throttle delay. Seems to be too simple, and yet it is, such a simple step to ensure the best possible quality of service for this type of traffic.