How would you tweak this Catalyst 2960 autoqos issue

[brettp]

I know a little bit about QoS, and by that, I mean basically nothing. I have set up autoqos (trust dscp) on our switches. Everything is okay for the most part, but I feel like the config is not optimal on a number of switches. For instance, queue 4 on this switch is seeing essentially zero traffic... So this means the buffers (reserved) are going to waste? What is the best way of adjusting this inefficiency? Tweaking the buffers (mls qos queue-set output 1 buffers 15 25 40 20) or lowering the reserved buffers on queue 4 and adding that percentage to the others? 

Queueset: 1
Queue : 1 2 3 4
----------------------------------------------
buffers : 15 25 40 20
threshold1: 100 125 100 60
threshold2: 100 125 100 150
reserved : 50 100 100 50
maximum : 200 400 400 200

output queues enqueued:
queue: threshold1 threshold2 threshold3
-----------------------------------------------
queue 0: 0 0 20842140
queue 1: 604463 3700139 2185570
queue 2: 0 0 44804247
queue 3: 0 0 0

mls qos srr-queue output dscp-map queue 4 threshold 1 8 9 11 13 15
mls qos srr-queue output dscp-map queue 4 threshold 2 10 12 14

[chrihussey]

Hello,

It has been quite some time since I've dug into the intricacies of the queuing, mapping and thresholds on the 2960. Admittedly it can be a bit involved and confusing (to me anyway) and I've forgotten most of it. Regardless, I just want to offer this suggestion. Unless you are having problems (voice quality, excessive packet loss, etc), I wouldn't change a thing. The fact that queue 4 is empty, only means that the switch isn't seeing those DSCP markings, so nothing gets put in that queue. This isn't necessarily robbing resources, and if any, it's negligible. Plus at some point there may be traffic with those markings, so it needs to be accounted for and according to the auto-qos it is allocated much less bandwidth anyway.

I guess I'm saying "If it ain't broke......."

You may disagree and still want to change it, so below is a link that goes into the specifics, and if you don't know already the following commands are useful in seeing how things are going:

"clear counters" (Just to get a clean starting point)

"clear mls qos interface statistics" (Just to get a clean starting point)

"sh mls qos interface g1/0/43 statistics"

Here's the link:

https://www.cisco.com/c/en/us/td/docs/switches/lan/catalyst2960x/software/15-0_2_EX/qos/configuration_guide/b_qos_152ex_2960-x_cg/b_qos_152ex_2960-x_cg_chapter_010.html#ID5926

Hope this helps.

[brettp]

Thanks for the input. While agree with “If it ain’t broke don’t fix it” – I think this is “kind of broke” in that the other queues are dropping stuff, if the additional buffers were made available to the other queues, not as many packets would be dropping. Ok, all of this is negligible because people aren’t complaining, but the potential is there. I’m just not sure of the best way would be… tweaking buffers or reserved buffers… or just something else all together.

[chrihussey]

Understood. I'd start be clearing all counters, monitoring the ports and traffic to better gauge what is going on. There may be some big talkers, physical issues, congestion points, etc. QoS, even if optimally implemented, doesn't solve all issues. You could spend a lot of time tweaking buffers, thresholds, etc. and still see the same results.

Regards