QoS on a Catlayst 6500

[wassim.kreidy]

Hi,

first I just want to say that I have never seen something as complicated to configure as the QoS on a Cat6k.

I don't know if you all agree with me, but if you don't please give me the "trick" to see it as easy and understandable as you do :-)

I have a 6509 running in hybrid mode (SUP ENGII CatOS/MSFC IOS) and I was looking into the QoS configuration guide for Cat OS 7.1. I understood that there are several types of queuing per port (which also depends on the port hardware).

I also tried to take a look into the IOS configuration guide for QoS (for Catalyst 6k running in IOS native mode) and found out that there are different types of queueing that for CatOS.

IS it true or am I mixing up stuff?

Is it a hardware related issue or a software one?

can any one point me out to any reference (not necessarely on CCO; a book may be) that explains QoS on the Cat6k in a clearer manner that the configuration guide?

Thank you for any help.

Ciao :-)

[ccie6145]

I agree 1000% it is way to complicated. Probably because cisco is still trying figure out the whole routing switch thing to some degree.

It becomes even more interesting when you add in Native Mode IOS, which is what I am running.

Ofcourse it depends on the application for QOS that you are trying to configure.

I recently was told to Rate Limit 1 departments internet traffic to 12 meg.

So I created an access list sourcing on that network, then a Class map pointing to that access list, then a policy map pointing to that class map and specifying 12 meg rate aggregate for the switch.

Which if you read doesn't really mean aggregate because I have DFC cards that track per blade.

Here is my config if you are interested.

class-map match-all dnr12c

match access-group 155

!

!

policy-map dnr12a

class dnr12c

police 12000000 375000 375000 conform-action transmit exceed-action drop

!

mls qos

interface GigabitEthernet1/2

service-policy input dnr12a

!

access-list 155 permit ip 10.5.0.0 0.0.255.255 any

Basically Aggregate policing is switch wide, or card wide if you have DFC cards. and Micro policing is per layer 4 flow, so if you setup a 100k rate for http no single PC connected to any single web server would transfer more than 100k on a Micro flow. Aggregate would be All HTTP traffic Switch wide. or all traffic from this segment.

Once you decide on your que, next choose your rate. don't worry about the Bucket thing, I just let the switch choose the default setting like this :

police 12000000

and it output this in the config

police 12000000 375000 375000 conform-action transmit exceed-action drop

It seems to be working for me. You were kinda vague on what kinda QOS you wanted to know about I hope this helps.

[bsivasub]

Wassim

Configuration guide for qos on cat6k needs to be read twice atleast so that you understand not only qos terminology but also the platform specific features.

Anyway, here is some tech tips on CCO which is much more easier to understand as it addresses each topic differently

Policing (native/hybrid)

http://www.cisco.com/warp/customer/473/102.html

Output scheduling (native)

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

Output scheduling (hybrid)

http://www.cisco.com/warp/customer/473/60.html

Queueing is hardware dependent. You have different commands for hybrid and native though. Interesting point with hybrid qos configuration is that you don't have to configure anything on the MSFC. everything is done (including aCLS ) at layer 2 supervisor (which has PFC ofcourse)

If you have any specific questions please post back or open a case with TAC.