Parent and child policy maps with LLQ

[BVC]

I'm trying to configure LLQ with parent and child policies on my two 4221 routers and I'm having trouble in understanding how LLQ works with the different policies so I'm hoping someone could help me with answering a few questions. Pretty much I've got a gig interface that I want to overall limit 2 to Mbps, this interface is split into two different sub interfaces on two different VRFs. I want the main data/VoIP interface to have a larger amount of bandwidth than the other VRF/sub interface. So, I've configured a parent policy interface for VRF A that shapes the bandwidth to 300000 bits, and the parent policy for VRF B shapes data to 1797120 bits, both of these policies overall add up to just under 2 Mbps. I've configured LLQ for the child policies that are linked to the parent policy, using both bandwidth and priority commands. All of this is for testing and not on production, each router is directly connected to each other. Policy configs are attached. 

1. When does LLQ kick in? Is it when each individual parent policy reaches their limit and it starts to reduce bandwidth based on bandwidth and the priority commands in the child policy, or is it when both parent policies on both sub interfaces reach their limits and that is when both child policies start to reduce bandwidth based on bandwidth and priority commands?
2. Do I need to configure another policy to put onto the actual physical interface itself to limit data (e.g. int gig 0/0/0)? Or is the shaping in the parent policy enough to keep the bandwidth under the 2 Mbps. 
3. When I use Iperf3 to test the bandwidth I don't get the results I'm hoping for, the admin data should be shaped down to 1155 Kbps when there is congestion, but it stays around at 1350 Kbps, way above the bandwidth limit. This worries me that LLQ isn't working as I hope. 

Any help will be much appreciated, thank you

[Joseph W. Doherty · Accepted solution]

#1

"When does LLQ kick in?"

When packets are in its queue.

That might happen when you exceed the parent's shaper limit, and it starts to queue traffic.  This would be per parent policy.  BTW, bandwidth statement do not set a queue bandwidth limit.  I.e. a class with even 1% bandwidth allocation may use 100%.  What a bandwidth statement does is provide a minimum bandwidth allocation guarantee, but it can be more than what's specified for a minimum if all the bandwidth allocations, together, don't allocate 100%.  Actually, what's provided is the ratio of the allocations.  So, for example, if you just have two classes one with a bandwidth allocation of 25% and the other 50% the minimums are 1/3 and 2/3.

#2

If all your traffic is processed by the two policies, with shapers, that should keep the interface from using more than the aggregate allowed by the policies.

#3

Again, bandwidth statements do NOT cap bandwidth.