Cisco 2960-X QoS egress queues question

[Peter Nemec]

Hi all,

I have got a WS-C2960X-48TD-L and I have a question regarding QoS.

==============================

First, QoS is disable as you can see:

==============================

#show mls qos
QoS is disabled
QoS ip packet dscp rewrite is enabled

==============================

Second, how come I am seeing packets entering the queues when QoS is disabled, as per the below output ?

==============================

#show mls qos interface GigabitEthernet1/0/48 statistics
GigabitEthernet1/0/48 (All statistics are in packets)

dscp: incoming
-------------------------------

0 - 4 : 39692336 0 34 0 0
5 - 9 : 8 0 0 2 0
10 - 14 : 1340 0 0 0 0
15 - 19 : 0 4 0 0 0
20 - 24 : 0 0 0 0 0
25 - 29 : 0 5 0 0 0
30 - 34 : 0 0 0 0 0
35 - 39 : 0 0 0 0 0
40 - 44 : 2 0 0 0 0
45 - 49 : 0 1111 0 219072 0
50 - 54 : 0 0 0 0 0
55 - 59 : 0 0 0 0 0
60 - 64 : 0 0 0 0
dscp: outgoing
-------------------------------

0 - 4 : 86964347 0 0 0 0
5 - 9 : 0 0 0 0 0
10 - 14 : 0 0 0 0 0
15 - 19 : 0 0 0 0 0
20 - 24 : 0 0 0 0 0
25 - 29 : 0 0 0 0 0
30 - 34 : 0 0 0 0 0
35 - 39 : 0 0 0 0 0
40 - 44 : 0 0 0 0 0
45 - 49 : 0 0 0 665587 0
50 - 54 : 0 0 0 0 0
55 - 59 : 0 0 0 0 0
60 - 64 : 0 0 0 0
cos: incoming
-------------------------------

0 - 4 : 216776891 0 0 5218 0
5 - 7 : 0 3764337 12136634
cos: outgoing
-------------------------------

0 - 4 : 87122268 118 0 111 59
5 - 7 : 0 665587 1024846
output queues enqueued:
queue: threshold1 threshold2 threshold3
-----------------------------------------------
queue 0: 0 0 0
queue 1: 173082 3142121 1691348
queue 2: 0 0 0
queue 3: 0 0 83806438

output queues dropped:
queue: threshold1 threshold2 threshold3
-----------------------------------------------
queue 0: 0 0 0
queue 1: 0 0 0
queue 2: 0 0 0
queue 3: 0 0 0

Policer: Inprofile: 0 OutofProfile: 0

==============================

I thought that QoS queues are only used when 1) QoS is enabled, and when there is a congestion and egress forwarding is  oversubscribed. 

But in this 2960-X architecture, the QoS egress queues are receiving frames when QoS is disables and there is no congestion. Is this something to do with the fact that 2960-X does QoS in hardware ?

==============================

Regards

Pete

[Joseph W. Doherty]

I thought that QoS queues are only used when 1) QoS is enabled, and when there is a congestion and egress forwarding is oversubscribed.

First, yes, my understanding too is the QoS egress queues (four provided) should only be used when QoS is enabled.  Of course, w/o QoS enabled, there's still an egress queue.  Don't know how QoS interfaces stats might, or might not, show stats for the non-QoS egress queue.  (It also possible you might be dealing with a "software defect" in how "stats" are accumulated and/or displayed.  If so, actual frame forwarding might be 100% correct.)

Also, BTW, I recall those stats are only cleared when the device is reloaded.  Are they incrementing, if not, possible they were generated at some time, in the past, when QoS was enabled?

Second, we may have different meanings about "congestion", and, also, queuing can still happen when there's no over subscription.

If you're wondering how for the latter (queuing w/o over subscription), consider a 100 Mbps egress with ten 10 Mbps sending traffic to it.  If ten frames are received at the same time, nine will need to be queued.  However, all ten frame can be transmitted before any more arrive, i.e. queuing but egress isn't over subscribed.

[Peter Nemec]

Hello Joseph,

thank you for your time and your response. I understand what you explained and that queuing does need to happen in normal operations when there's no over subscription, and there are egress buffers (queues) to deal with that, in 2960-X it's 4MB shared buffer.

When I ran "show mls qos interface GigabitEthernet1/0/48 statistics" multiple times, yes, I do see packets increasing in queue 1 and queue 3.  I also did not find a command to clear these counters, and those would clear upon reload.

When I look at queue-set 1 for example (see below), I see that all queues have 25% of buffer allocated.

=========================================

#show mls qos queue-set
Queueset: 1
Queue : 1 2 3 4
----------------------------------------------
buffers : 25 25 25 25
threshold1: 100 200 100 100
threshold2: 100 200 100 100
reserved : 50 50 50 50
maximum : 400 400 400 400

=========================================

My ultimate question that I wanted to understand is that if these 4x queues are used for egress transmission, and if only queue 1 and queue 3 are receiving packets as per my example with port "GigabitEthernet1/0/48", shouldn't we de-allocate buffer space from queues that are not used, and re-allocate those to queue 1 and queue 3 ?

(config)#mls qos queue-set output 1 buffers 0 50 0 50

Would that make any difference with regards to having more buffer % in the queues that are used  and theoretically, this interface would experience less output drops if a congestion was to occur because it would have larger buffers for egress  transmission?

Regards,

Peter 

[Joseph W. Doherty]

". . . yes, I do see packets increasing in queue 1 and queue 3."

Well, again, if QoS is disabled, to me, this would appear to be an "issue".  Most likely "cosmetic" as over the years, at least on the somewhat similar 3560/3750 series saw various software defect concerning some of the interface stats.

"When I look at queue-set 1 for example (see below), I see that all queues have 25% of buffer allocated. "

Yup, that's the default.

"My ultimate question that I wanted to understand is that if these 4x queues are used for egress transmission, and if only queue 1 and queue 3 are receiving packets as per my example with port "GigabitEthernet1/0/48", shouldn't we de-allocate buffer space from queues that are not used, and re-allocate those to queue 1 and queue 3 ? "

If we were using QoS (which you're not), and seeing actual drops (which you're not - you stats only showed enqueued frames), maybe.  Why "maybe"?  Because normally I might adjust WTD 1 and WTD 2 first and/or vary buffer reservation percentage between common pool and interface reserved buffers and/or increase total amount of buffers an interface might acquire from common pool.

[Peter Nemec]

OK great Joseph, again thx for the reply, one last thing I am struggling to understand, below are the defaults for WTD 1 and WTD 2 etc...

==========================================

h2-15-1e.ord1#show mls qos queue-set
Queueset: 1
Queue : 1 2 3 4
----------------------------------------------
buffers : 25 25 25 25
threshold1: 100 200 100 100
threshold2: 100 200 100 100
reserved : 50 50 50 50
maximum : 400 400 400 400
==========================================

In cisco's official documentation, this is how they explain the  WTD 1 and WTD 2 values:

• For drop-threshold1 drop-threshold2, specify the two WTD thresholds expressed as a percentage of the queue’s allocated memory. The range is 1 to 3200 percent.

I am pulling my hair out as to what exactly this means, say if I configure something like this (this is just a random example so I can ask the question)

"mls qos queue-set output 1 threshold 2 100 3200 50 800"

What on earth would the 3200% actually mean ? Percentage of what exactly ? The documentation says that "percentage of the queue’s allocated memory", where in my example the allocated memory is 25% right ? So the way I am looking at this right now, and I am obviously probably totally wrong is that I configured 3200% of 25% ?? Which just does not make logical sense, because I must be looking at this totally wrong.

Regards,

Pet

[Joseph W. Doherty · Accepted solution]

Reading the "normal" 2960/3560/3750 QoS documentation, left me unable to understand how these device QoS parameters interacted.

Fortunately, a "document" ()https://community.cisco.com/t5/networking-documents/egress-qos/ta-p/3122802) was published on this site to much better explain QoS parameters.  That said, it may take a couple of readings to understand it.