Solved: BGP, PAK-Priority, and QoS

prakkdangc · ‎12-15-2009

Hi,

I have heard that IGP packets on the Cisco control plane automatically get marked with CS6 and are escorted to the tx-ring with high priority. BGP supposedly does not get this special treatment, but instead gets TCP to mark the packets with CS6.

A couple questions:

Do I never have to mark the packets with CS6 if the IOS/Router is doing it automatically (assuming an all-Cisco network)?
Is there one queueing strategy that is better than another if I am running both IGP and EGP protocols at the same time?

Thanks

Dave

Giuseppe Larosa · ‎12-16-2009

Hello Dave,

BGP packets are marked with CS6 you can verify this with a packet capture so I don't see any difference with IGPs.

Depending on the platform you are using and the IOS image it is running, an hidden system queue that takes care of routing protocol messages may exist or not in your device.

You don't need to mark the routing protocol packets, however in platforms that miss the hidden system queue you may need to create a class map for this traffic to provide enough resources or neighborships may flap under heavy load conditions on the link.

C3600 and ISR should have this system queue something changes in very recent IOS 12.4T releases

On high end routers like GSR, C7500 you need an explicit class for routing protocol traffic.

Hope to help

Giuseppe

View solution in original post

Giuseppe Larosa · ‎12-16-2009

Hello Dave,

unfortunately the effect of using 100% max-reserved-bandwidth can be that of depleting the system hidden queue.

see

http://www.cisco.com/en/US/docs/ios/qos/command/reference/qos_m1.html#wp1039174

>> If you do use the max-reserved-bandwidth command, make sure that not too much bandwidth is taken away from best-effort and control traffic.

There is a note already pointed out by Edison Ortiz in another thread, that the command is no longer supported in releases after 12.4(20)T that implements Hierarchical QoS framework.

Depending on the link speed I think you can use 90 or 95 percent but not 100%

I'm afraid that my first post can be useful but it is not complete!

Hope to help

Giuseppe

View solution in original post

Giuseppe Larosa · ‎12-16-2009

Hello Dave,

BGP packets are marked with CS6 you can verify this with a packet capture so I don't see any difference with IGPs.

Depending on the platform you are using and the IOS image it is running, an hidden system queue that takes care of routing protocol messages may exist or not in your device.

You don't need to mark the routing protocol packets, however in platforms that miss the hidden system queue you may need to create a class map for this traffic to provide enough resources or neighborships may flap under heavy load conditions on the link.

C3600 and ISR should have this system queue something changes in very recent IOS 12.4T releases

On high end routers like GSR, C7500 you need an explicit class for routing protocol traffic.

Hope to help

Giuseppe

prakkdangc · ‎12-16-2009

Thanks Giuseppe! I feel much better now.

Dave

marikakis · ‎12-16-2009

Yeah Giuseppe, good post!

prakkdangc · ‎12-16-2009

Hey Guys,

I just had a thought. If you configure max-reserved-bandwidth 100 on an interface, will you affect the "hidden" queue and it's ability to do PAK prioritization for routing protocols?

Dave

Giuseppe Larosa · ‎12-16-2009

Hello Dave,

unfortunately the effect of using 100% max-reserved-bandwidth can be that of depleting the system hidden queue.

see

http://www.cisco.com/en/US/docs/ios/qos/command/reference/qos_m1.html#wp1039174

>> If you do use the max-reserved-bandwidth command, make sure that not too much bandwidth is taken away from best-effort and control traffic.

There is a note already pointed out by Edison Ortiz in another thread, that the command is no longer supported in releases after 12.4(20)T that implements Hierarchical QoS framework.

Depending on the link speed I think you can use 90 or 95 percent but not 100%

I'm afraid that my first post can be useful but it is not complete!

Hope to help

Giuseppe

prakkdangc · ‎12-16-2009

That's good news Giuseppe. Thanks for the follow up.

Dave

fly · ‎10-09-2015

mark