Different values for Eth and cbQoS

[Bartman2002]

Hi all,

I configured QoS on a lab router, when I compare Eth and QoS (Pre & Post) values, the QoS values are slightly higher. (20%). When I deal with misconfiguration I supposed to see more traffic on Eth.... Can anyone give me the answer?

Bart

[pauld]

Hello Bart,

I don't fully understand your question, so I've got a few questions for you.

I configured QoS on a lab router, when I compare Eth and QoS (Pre & Post) values, the QoS values are slightly higher. (20%). When I deal with misconfiguration I supposed to see more traffic on Eth.... Can anyone give me the answer?

Can you clarify what you mean by comparing Eth and QoS? What is Eth? Do you mean port speed?

The QoS values are slightly higher than what? From my understanding, Quality of Service is a set of bits that determine the priority of the traffic to ensure quality and reliability in conversations across the network, so I'm unsure how this can be 20% higher than what you're expecting.

Can you post a screen shot of what you're looking at?

Thanks,
Paul

[Bartman2002]

Hi Paul

I made some screencaptures and included them as attachment. The cbQoS configuration is for outbound. See the diff. 13 / 18 K. I don't understand why the QoS value higher compared to the Eth.

[pauld]

Hi Bart,

What interfaces are your CBQoS policy applied to? The CBQoS Bytes Pre/Post report is showing how the policy matched for all the interfaces it's applied to, not just the Fa0/0 interface. It's recommended to configure CBQoS on a per-interface basis for the best QoS.

Thanks,
Paul

[Bartman2002]

Hi Paul,

Thank you for brainsharing....

cbQoS is only applied to Fa0/0

Is it possible caused by header information??

interface FastEthernet0/0
ip address 10.0.0.237 255.255.252.0
ip nbar protocol-discovery
ip flow ingress
ip flow egress
duplex full
speed 100
service-policy output Bart
!
interface FastEthernet0/1
description $ES_LAN$
ip address 192.168.1.1 255.255.255.0
ip nbar protocol-discovery
duplex full
speed 100

Bart

[tomp]

Hi Bart,

Can you show me your service policy and class map definitions?

What IOS version are working with?

- Tom

[Bartman2002]

Hi Tom,

It's quite basic...


Cisco IOS Software, 2800 Software (C2800NM-ADVIPSERVICESK9-M), Version 12.4(24)T4, RELEASE SOFTWARE (fc2)


class-map match-any Bart
match any

policy-map Bart
class Bart

interface FastEthernet0/0
ip address 10.0.0.237 255.255.252.0
ip nbar protocol-discovery
ip flow ingress
ip flow egress
duplex full
speed 100
service-policy output Bart

[tomp]

I'm starting to think you may be experiencing a Cisco bug, but I'm verifying in my lab right now.

I'm working with:

Cisco IOS Software, 2800 Software (C2800NM-ADVENTERPRISEK9-M), Version 15.1(3)T, RELEASE SOFTWARE (fc1).

I'll keep you posted.

[Bartman2002]

Hi Tom,

I upgraded my router to the latest 15.1(3)T and still experience the same mismatch....

Bart

[tomp]

Hi Bart,

I just verified that I'm also seeing a difference. Mine is closer to 100k off.

I'm going to contact TAC and see what they have to say.

- Tom

[Bartman2002]

Looking forward to their reply.

[tomp]

Hi Bart,

I haven't forgot about you. I have a TAC case open and I've been trying to connect with them.

I just viewed my trending graphs and they are now showing accurate information (withing a couple k of each other)

As you can see here if you add up all of the cbQoS and compare with the outbound port utilization, they are very close.

Are your reports still tracking apart?

cbQoS.png (172.68 KiB) Viewed 14954 times

[tomp]

Hi Bart,

I have a hunch why these values are slightly different.

When we poll for cbQoS values, we grab the counter and report exactly on the counter value. When we poll for Port Utilization we poll for the counter value then divide by 300, which then makes this value an average and not an exact figure. We do this because the poller polls every 5 minutes or 300 seconds and the counter associated to port utilization is a counter that increases over time, not a gauge that reports the amount of utilization at a specific moment.

I think that Pre (1.3.6.1.4.1.9.9.166.1.15.1.1.5.$INSTANCE)/Post Policy (1.3.6.1.4.1.9.9.166.1.15.1.1.9.$INSTANCE) traffic is a gauge. We will see what TAC says, but I have a feeling this is the answer.

- Tom

[tomp]

Hi Bart,

Cisco has not gotten back to me on this. They have been unable to confirm/deny this is a bug.

- Tom