Understanding this QoS setup

[Colin Higgins]

I've been looking through the documentation for QoS and Avaya phones, and I see the following recommendations for setup on a 3750E

mls qos map cos-dscp 0 8 16 26 32 46 48 56

mls qos srr-queue input cos-map queue 2 threshold 2 4 6 7

mls qos srr-queue input dscp-map queue 2 threshold 2 33 34 35 36 37 38 39 48

On a trunk uplink to router:

interface GigabitEthernet1/0/24

switchport trunk encapsulation dot1q

switchport trunk allowed vlan 124-126

switchport mode trunk

srr-queue bandwidth share 10 10 60 20

srr-queue bandwidth shape 10 0 0 0

queue-set 2

mls qos cos 6

mls qos trust cos

auto qos voip trust

no cdp enable

spanning-tree portfast

OK, so correct me if I am wrong here, but we are

1. Taking packets with cos value 6 (as Avaya marks them), and putting them into input queue 2 threshold 2

2. We map the dscp value (48) to the queues

3. On every cycle 10 packets are serviced from queue 2, 60 from queue 3, and 20 from queue 4

But here is where it gets a bit confusing

Why is the traffic being put into queue 2 and not queue 1 as in other QoS implementations?

Is queue 2 the priority queue by default (as I have read)? If so, which queue do we use, 1 or 2?

Why is there no priority-queue out statement on the trunk?

Why is the srr-queue bandwidth shape set to 0 for queue 2? Shouldn't that be 10?

thanks!

[JohnTylerPearce]

interface GigabitEthernet1/0/24

switchport trunk encapsulation dot1q

switchport trunk allowed vlan 124-126

switchport mode trunk

srr-queue bandwidth share 10 10 60 20

srr-queue bandwidth shape 10 0 0 0

queue-set 2

mls qos cos 6

mls qos trust cos

auto qos voip trust

no cdp enable

spanning-tree portfast

I'm going to highlight several items for you.

srr-queue bandwidth share 10 10 60 20

srr-queue bandwidth shape 10 0 0 0

Notice with the above, that you are shaping queue 1. So you will want to look at the "srr-queue bandwidth share 10 10 60 20", as "0 10 60 20"

mls qos cos 6

mls qos trust cos

With "mls qos cos 6", you are tagging EVERYTHING that is being received by that port as cos 6. Using "mls qos trust cos', will trust the CoS values coming in from any device on the other side of the link.

So by using the "mls qos cos 6" you have essentially makde the "mls qos trust CoS" pointless. I would take out the command "mls qos cos 6", unless you need that there for a specific reason.

Also, you are assigning queue-set 2, to that port, so I don't know what specific values you have for buffers, thresholds, etc.

I hope this helps you out some. Also, I do not have any experiencing setting up those IP Phones, so take my advice as you see fit.

Have a good one,

** Update**

I forgot the part about, why there isn't a "priority-queue out". This has not been configured, so that's why it isn't there. If you insert this command, it will service that queue untill it's full, then the other 3 queues will be able to be serviced.

[Colin Higgins]

I guess what I am really wondering is why they are putting voice traffic into queue 2 vs. queue 1, and why they are shaping queue 1 and not 2

[JohnTylerPearce · Accepted solution]

Ok, so this port is a gigabit port 1Gbps. So, the 'queue-set 2' command is assigned that port to queue 2. Now, since I don't have access to the switch (nor should I), I don't know what the values are for the 4 buffers, 2 threshold values, bandwidth, and priority for queue-set 2. You can run 'show mls qos queue-set 2' and it should give you those parameters.

srr-queue bandwidth share 10 10 60 20

srr-queue bandwidth shape 10 0 0 0

My math could be wrong, but having 'srr-queue bandwidth shape 10 0 0 0' is taking 1/10 which equa's 100Mbps, is being shaped, on queue 1. Remember you can have two "queue-sets" ingress queues.

So since you have queue 1 being shaped, think of 'srr-queue bandwidth share 0 10/(10+60+20) | 60/(10+60+20) | 20/(10+60+20)

So, at your current configuration, it looks like, traffic marked as CoS 6 (Which is everything entering your port right now, regardless if it's PC and or VoIP traffic), is going to CoS 6, and both ingress and egress queueing, will use that to map that CoS value to it's Q#T#. If you run 'show mls qos maps cos-dscp' you will see what DSCP value CoS 6 maps too.

Basically, on a 1000Mbps link, 100mbps is placed onto the shared bus/ring, from Queue 1 on Queue-set 2, and Queues 2-4 have 900mbps, between them.

[Colin Higgins]

OK, I think I have this

If you use priority-queue out, it will override all shaping and sharing --thus it is not enabled here.

Queue 2 is the priority queue by default (or at least from what I have read)

So in our example, this traffic should be in queue 2 threshold 3

mls qos srr-queue input cos-map queue 2 threshold 3  3 5 6

[JohnTylerPearce]

Colin,

The command 'mls qos srr-queue input cos-map queue 2 threshold 3 3 5 6' will map CoS values 3,5, and 6 to threshold 3, which by default will only drop traffic, if the queue is 100% full.

Glad, I could help you out!