Class Map Question

[Adam Frederick]

I have a question regarding using class maps for QoS. I am using LLQ on my Internet connection, with a simple config, I need to guarantee 2 hosts 768 of my 1544 T1. I've set it up as follows;

class-map Host1

match access-group 50

!

class-map Host2

match access-group 51

!

access-list 50 permit 10.10.10.1

access-list 51 permit 10.10.20.1

!

policy-map QoS

class Host1

priority 384

!

class Host2

priority 384

!

int s0

service-policy output QoS

!

And finally, the question:

When I do a show int s0, the output says I have 390k of free bandwidth. Can someone help clear this up? The way I understand, the QoS shouldn't even take effect until the link is congested, so do I not have the 1544 available at all times? I'm just afraid I'm starving bandwidth. When I add 384x2+390 I get 1158 (far from 1544). Any insight on this would be greatly appreciated.

[pkhatri]

Hi,

The maximum bandwidth available on an interface to be assigned to qos classes is by default, 75% of the interface bandwidth. In your case, this is 0.75*1544, which is how you get the 1158k. You can change this 75% figure by using the max-reserved-bandwidth command.

When you assign a service-policy to an interface, the bandwidth assigned to the various classes is subtracted from this available bandwidth.

The available bandwidth shown in the output of show interface is how much more bandwidth is available to be assigned to allocated to qos classes. This is a static bandwidth figure, and does not reflect the actual bandwidth on the interface.

Hope that helps,

Paresh

PS. Pls rate helpfulposts.

[pkhatri]

Hi again,

Just a bit more info to supplement my last post.

The default figure of 75% for max-reserved-bandwidth allows 25% for the following:

- layer 2 keepalives etc

- traffic not matching any of the user-defined classes i.e. traffic that falls into the class-default class.

For an interface with a bandwidth of 1544k, you could allocate a service-policy with classes that assign a total bandwidth of up to 1158k i.e. the total of bandwidth assigned using the 'bandwidth' and 'priority' commands can be up to 1158k.

Hope that helps,

Paresh

[mheusinger]

Hi,

just to make sure: the "available" bandwidth is only to calculate what amount of bandwidth you can give to classes in your policy. The reasonable approach is to leave some bandwidth for traffic not accounted for in your policy-map, like Layer2 keepalives, routing updates and so on as already stated in the previous posts.

The whole Queueing configuration only kicks in when there is an overload situation. So to say your policy-map is a ruleset for the router what to do with IP packets which can not be sent immediately because the interface is congested. You basically define in which order the IP packets have to be taken out of packet memory and sent over the interface.

In case there is no congestion any traffic can occupy all your bandwidth, i.e. you always have T1.

Another thing is: with your "priority statement you define also a policer. So in case there is an overload situation you drop excess traffic from the two hosts to limit them to 384 kbps.

In case you do not want to limit them, but to give a minimum bandwith guarantee, you should use "bandwidth" instead of "priority" statements.

Hope this helps

Martin