QoS confusion

[Colin Higgins]

I am reading through some of the documentation for WRR queuing, and I am la little confused on a couple things:

Each switch has a certain number of egress queues, and one of these is the priority queue. So a 3750 will be a 1p3q1t (XpYqZt), meaning that the priority queue is listed first here.

But during configuration of WRR, the documentation from Cisco states "The queues are always numbered starting with the lowest-priority queue possible, and ending with the strict-priority queue. In their example they use : queue 1 will be the low-priority WRR queue, queue 2 will be the high-priority queue, and queue n will be the strict-priority queue.

So which one is the priority queue? Is the 1p3q1t backwards? Shouldn't the priority queue be at the end?

If we do

wrr-queue bandwidth 50 (Q1)

Is the priority queue getting 50? Which queue is that?

[Jon Marshall]

Colin

You're confusing me now

To my knowledge the 3750 doesn't use WRR it uses SRR and the priority queue doesn't participate in SRR.

So are you perhaps referring to a 6500 switch or something similiar ?

Jon

[Colin Higgins]

Yes, you are correct

Let's pretend we are talking about a 6500 series switch here. How am I supposed to interpret the queue nomenclature?

[Jon Marshall]

colin-higgins wrote:
Yes, you are correct
Let's pretend we are talking about a 6500 series switch here. How am I supposed to interpret the queue nomenclature?

Okay, the thing to be aware of is that WRR queueing does not apply to the strict priority queue. It only applies to the standard queues. So with your example of a 1p3q1t architecture when you use the WRR command you are only applying this to the 3q part of the 1p3q1t. So in answer to your question when you use wrr-queue bandwidth 50 you are applying that to q1 from the 3q bit of 1p3q which is, as you say, the lowest priority queue.

The strict priority queue is always serviced as soon as a packet arrives in the queue so you do not configure a bandwidth allocation for it.

Jon

[Colin Higgins]

ok, it is starting to make sense

what about the 4th queue (as in the case of a 3750) or "1t" queue?

[Jon Marshall · Accepted solution]

colin-higgins wrote:
ok, it is starting to make sense
what about the 4th queue (as in the case of a 3750) or "1t" queue?

1t means 1 threshold ie. if the architecture supported 2t for example then in each standard queue you could configure 2 thresholds. This is useful for example if you map 2 CoS values to the same queue but you actually want to give preferential treatment to one of those values within the same queue. To do this you would allocate different thresholds to each CoS value and when the queue started dropping packets it would drop the less favoured CoS value first.

Not sure what you mean about 4th queue on 3750. As already said SRR doesn't apply to the strict priority queue on the 3750 in the same way WRR doesn't apply to the strict priority queue on the 6500.

Jon

[JoeKeegan3]

Yeah it's confusing.

Basically the 1p3q1t doesn't correlate with the actual number for the queues, it's just describing the capabilities of the card/switch.

Some switches will have their priority queue as queue one, such as the 3750, while others have other queues as their priority queues as three such as a 4500.

So you can't tell from the #p#q#t number which queue is the priority queue, you can just tell from that number how many priority queues (I think this is always 1 if it has any) and more importantly how many regular queues and thresholds the card has.

Joe

[Colin Higgins]

OK, I underastand now

Thanks for your help guys!