QoS Priorities

[David VR]

Hi there,

I have read and learned a lot concerning QoS.

What is still unclear to me is the prioritisation mechanism.
From what I understood is that there is a schedular servicing the queues.
Depending on the switch model IOS, XE-OS, NX-OS there is a different algorithm running.

Like DWRR, or WFQ, ...

What I don't understand is how do you tell which queue should be served first.
Normally you classify the interesting traffic, mark it with CoS, DSCP or qos-group, ...

In the policy you assign a given bandwidth per queue.

But how do you say this queue needs to be served before other queues?
In most switches you have a maximum of 8 queues, so why should you have like 56possibilities with DSCP if you can only map it to 8queueus?
Same with qos-groups you have more than 100 posibilities but how do you match these possibilities with your queues?


Thanks for the clarification.

David

[Joseph W. Doherty]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages wha2tsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

Perhaps you're cannot see the (QoS) forest for all the trees?

As you note, individual device QoS features vary much.  What any one device's QoS configuration actually does, depends on understanding the device's QoS capabilities.  However, understanding the intended overall QoS policy might not be possible from a device's QoS configuration because it may lack features to do all that's actually desired.

In most switches you have a maximum of 8 queues, so why should you have like 56possibilities with DSCP if you can only map it to 8queueus?

Yes, modern Cisco switches often support 8 queues, but older models often only supported 4 queues, and some really old Cisco devices only supported 2 queues.

BTW, DSCP recommendations don't really recommend a need for 56 queues.  (How did you get the value 56?)  Many DSCP settings are undefined or are for different drop settings.

What you really want to do, is logically define your QoS policy, and then determine how best it might be accomplished on any individual device.

[Furose M · Accepted solution]

But how do you say this queue needs to be served before other queues?

Furose: you have to explicitly tell the switch that this is a priority queue. you typically do it under "class-map" for whichever class you want it to be a priority queue. it doesnt matter which DSCP are being matched in that class. if you tell the switch that the traffic falling in this class should be serviced first, it will be.

In most switches you have a maximum of 8 queues, so why should you have like 56possibilities with DSCP if you can only map it to 8queueus?

Furose: not all the 56 DSCP values are used, however, the reason for having more than 8 DSCP values is, you can service traffic falling under same classes differently with different drop threshold values.

Same with qos-groups you have more than 100 posibilities but how do you match these possibilities with your queues?

Furose: you can match more than 1 qos-group to a single queue.

[David VR]

Thanks a lot for the clarification.

This makes sense.