QOS - default class question

[rll3]

#######sh policy-map

Policy Map VOICE-WAN

Class VOICE

Strict Priority

Bandwidth 500 (kbps) Burst 12500 (Bytes)

*** Class class-default

Flow based Fair Queueing

** Bandwidth 0 (kbps)

exponential weight 9

class min-threshold max-threshold mark-probablity

----------------------------------------------------------

0 - - 1/10

1 - - 1/10

2 - - 1/10

3 - - 1/10

4 - - 1/10

5 - - 1/10

6 - - 1/10

7 - - 1/10

rsvp - - 1/10

With this configuration, is the default working or would traffic outside the strict priority be sent best effort?

[Joseph W. Doherty]

I would expect it to work, and as to traffic not in the LLQ class, one could argue flow based FQ with RED, a form of AQM (active queue management), is not best effort.

[rll3]

I'm going to pose the question a bit differently;

What I'm trying to find out is how any traffic not in the Strict Priority would be sent... Would it be sent best effort or would it be sent as Class-Default even if the bandwidth is set to '0'?

What I need to understand is what happens to traffic that is not in the Strict Priority class/policy...

[lee.reade]

hI,

Unless you have configured max-reserved-bandwidth 100, the class class-default always has 25% of the available bandwidth.

If you had a 100MB link, then you can only assign 75% of that to any class you define, UNLESS you config the command above on the interface, in which case you can assign 100% to your class, and the class default gets nothing.

What this means that if you give your class 100% of the bandwidth, then it CAN send up to 100% of the link bandwidth, however if it does not then other classes, ie class default can use whats left.

If you start to get congestion then your class will use up to its 100% and the class default will get dropped as there is no bw available.

HTH

LR

[Joseph W. Doherty]

Assuming you haven't configured LLQ to be able to obtain all the link bandwidth, other classes, when there's congestion, will have minimally available to them the bandwidth between the LLQ max and link's max. So for instance, if you configured priority 500 Kbps on a T1, other classes would have minimally about 1 Mbps available to them since LLQ will police (drop) any priority traffic that exceeds 500 Kbps. If there are no other classes but the LLQ priority class and class default, class default could obtain all the excess bandwidth above the LLQ max or not being used by LLQ regardless of the bandwidth setting for class default.

PS:

Not sure you can configure a zero bandwidth, recall 8 Kbps might be a minimum on many IOSs.

[rll3]

Hi Jd,

Thanks for your explanation. It's starting to make sense a bit.

Okay, the way things worked out is that the LLQ is Strict Priority (500kbps). Of course the Class-default is automatically created, but there was never any bandwidth specified, so what is displayed is (0kbps).

The reason for my questions is that now we are experiencing drops on this link and i'm trying to figure out if the left over link bandwidth is being fully utilized or not.

T1 = 1.544mbps

Strict Priority (VoIP) = 500kbps

Class-Default = 0kbps (maybe 8kbps per IOS version)

Available Link BW = 658kbps

[1544kbps - 25%(386kbps) = 1158kbps - 500Kbps = 658kbps]

Now contending for this (658kbps) bw are an email, File, and Database server along, with a large system link, and a web/collaboration server.

It should be enough, but with the drops it may not be... there are also complaints of slow performance, so additional QOS is need but how far should we go with it is the question.

[Joseph W. Doherty]

If you just have LLQ and class-default, class-default should be able to use almost all the link's bandwidth not being used by LLQ.

To better see what's happening, use the "show policy-map interface output" command for stats. (You have applied the policy to an outbound service policy, correct?)

[rll3]

yes

I think it might be displayed in an earlier post.

[Joseph W. Doherty]

Looks like your original post has just "show policy-map", that doesn't show the interface stats.