QoS issues and Port-channel on Cat6513

[blagoy.popov]

Hi,

Could some body help me with this? I have two Cat6513 running Port-channel between them.

Does anybody know where should I apply the service-policy command? On the Physical interfaces ( Gigabitethernert ) or on the Portchannel interface? Also I found in the documentation that CBWFQ is not supported but the switch accepts command anyway. It looks though that the counters does not work. When I issue show policy-map interface all counters state zero (0). Another question, when I can apply wrr commands on the physical interfaces from which the portchannel is made and they are accepted, but I could not find the way to verify that they are working :( sh mls qos int does not give us much information.

In general any info will be appreciated. Thanks a lot in advance.

Best Regards, Blago

[owillins]

There is no equivalent to CBWFQ on a 6500. Although the commands might be accepted I don't think it is supported. Here is a tech tip that discusses the QoS features supported with 6500 and also WRR.

http://www.cisco.com/warp/public/473/73.html

[blagoy.popov]

Hi,

Thanks a lot for the answer. I had red this. There is one thing that bothers me.Should I apply the wrr commands on the phisycal interface (for example: int gi 0/1 ) or I should apply them on the portchannel interface.

Let me show you what I mean:

When I issue:

show interface gigabitethernet 6/2 capabilities

I get this:

QOS scheduling: rx-(1p1q4t), tx-(1p2q2t)

But when I issue

show interface portchannel 101 capabilities

I get something different

QOS scheduling: tx-(1q4t)

Which one should I consider? And where I should apply wrr commands ? Does this tx-(1q4t) mean that a portchannel interface might have some kind of software buffers or what ... ?

Thanks a lot.

[Prashanth Krishnappa]

Just because you see commands in IOS parser does not mean it is supported. Anyways, you will need to apply any supported QoS commands on the port-channel interfaces rather than physical ports. Make sure the Etherchannel members have same QoS capabilities.

[blagoy.popov]

Thanks a lot for the answer.

So I have 1 queue with 4 treshholds on the port-channels? Does it matter if the Etherchannel members are from different DFC modules, or all they need is to have the same QoS capabilities? Is there any other way to observe the effect of the QoS commands applied instead of:

show queueing interface

show mls qos ip interface .

Can you advise about method for verifying the deployed QoS techniques on gigabit interfaces? Maybe with desktops with gigabit ethernet cards and traffic generators?

Thanks a lot in advance .

[blagoy.popov]

So anybody?

Could somebody provide me with some more information?

So I might did not explain well. When I issue show interface portchannel 111 capabilities

I get that capabilities for QoS on the port channel are 1q4t(1 queue and 4 thresholds) if a issue the

show interface giga 4/1 capabilities ( member of the port-channel) a get that the physical interface has 1p2q2t capabilities. I got confused because as long as I know it is not possible to have congestion on a software interface ( like portchannel ) so it sound more logical to me to make queuing on a physical interface. Also the software interface does not have any hardware buffers. So, where the queuing is done. Anyway to have successfully QoS queuing deployed what should I issue? Something like this:

Interface portchannel xxx

wrr-queue random-detect min-threshold 1 1 5 70 95

wrr-queue random-detect max-threshold 1 5 70 95 100

wrr-queue cos-map 1 1 1

wrr-queue cos-map 1 2 0 2

wrr-queue cos-map 1 3 3 4

wrr-queue cos-map 1 4 5 6 7

Or I should apply something like this:

interface range GigabitEthernet4/1 - 2

wrr-queue queue-limit 60 20

wrr-queue bandwidth 70 30

! Sets the WRR weights for 70:30 (Q1:Q2) bandwidth servicing

wrr-queue random-detect min-threshold 1 1 5

wrr-queue random-detect max-threshold 1 60 100

wrr-queue random-detect min-threshold 2 40 80

wrr-queue random-detect max-threshold 2 80 100

wrr-queue cos-map 1 1 1

wrr-queue cos-map 1 2 0 2

wrr-queue cos-map 2 1 3 4

wrr-queue cos-map 2 2 6 7

Any suggestions and/or explanations will be greatly appreciated.

Best regards

[ruben.montes]

Hello,

have you arrived to any conclusion? I'm facing the same problem as you with etherchannels...

Thanks,

Ruben

[mmelbourne]

I believe the wrr-queue commands need to be applied to the physical interfaces of the EtherChannel, not the port-channel logical interface. In fact, using 12.2(18)SXD3 on the Sup720s, I don't have the wrr-queue command available under the port-channel interface.

How did you determine your wrr queue-limit sizes and buffer weights? If most of our traffic is "Best-Effort" (CoS=0), but we implementing a small scale IPT solution where voip-bearer and voip-control traffic needs to be queued, how should the queue sizes and weights be determined?

Do the buffer sizes only come into operation at times of congestion? For examples, can a single queue use 100% of the queue size when no other traffic is competing for bandwidth?

[blagoy.popov]

Hi everybody,

Yes you are right.All queueing must be done on physical interface. This is what I finally got from Cisco.

" Configure transmit queue configuration on the physical interfaces. The configurations must be equal on all ports participatig.

No software checks to validate that the transmit queue configurations on physical ports are same"

For the wrr queue-limit sizes and buffer weights you should have an idea what your traffic looks like or so called traffic patern. You should know approximately what is the amount of the different traffic types. And once you deploy the queues you should observe the result and adjust the queues if there is a need.

Check this url:(you will need cco login)

http://www.cisco.com/warp/customer/473/73.html

Hope this helps.

Best regards, Blagoy Popov

[mmelbourne]

A very useful link. Many thanks.