WAN Bandwidth Allocation in policy map applied to WAN interface

[TyrosAdmin]

his might be more of a question for QOS, but it does have to do with voice, so I'll give it a shot here.    We're a callmanager 7 environment using skinny and mcgp.

I have a 4x T1 MLPPP WAN solution at my main site, and in case I decide to increase my bandwidth, I want to be able to keep my same bandwidth allocations for tiered voice and data.

My WAN provider will automatically set the QOS on the links the way I have it set up, but if packets are to drop, I'd rather them drop at my router than against the wan where I have zero visibility.

I want my allocations to be in 4 tiers.    Top Tier gets 40%, and the other three tiers get 20% each   Class1 is for voice, Class2 is for VOIP Control/MCGP/Skinny, Class3 and Default are for remaining traffic.

I did some research, and tried this config.   However, what I got on the Class3 was an error that after setting priority percent 40 and 20 on the upper two classes, that there was only 15% of bandwidth remaining.   Last I checked, 40 and 20 equaled 60, not 85.   There must be something I'm missing.

Here's the policy map I'm wanting to set for my WAN traffic.

policy-map DV-L3-OUT

  class Class1

   priority percent 40

   set dscp 46

  class Class2

   priority percent 20

   set dscp 32

  class Class3

   priority percent 20

  class class-default

   fair-queue

I would appreciate any feedback that anyone might be able to offer.

James

[Edison Ortiz]

Please refer to the max-reserved-bandwidth command.

http://www.cisco.com/en/US/docs/ios/qos/command/reference/qos_m1.html#wp1054626

I don't recommend assigning a priority queue to traffic class other than voice bearer.

I also don't recommend providing for more than 33% for PQ in general as it will starve other queues.

All other classes should use CBWFQ.

Regards,

Edison

[skarthic]

Hi James,

While configuring priority queuing/ LLQ  in an interface, we can allocate a maximum of 75% of the interface bandwidth for the priority traffic

as per your classification.

The rest 25% of the BW will be unused for any data traffic, rather will be used for the overhead traffic like routing updates etc.

So while configuring priority you can configure only upto a max of 75%.

I have attached a link that explains the configuration of priority traffic in a interface:

http://www.cisco.com/en/US/tech/tk543/tk757/technologies_tech_note09186a0080103eae.shtml

Regards.

Please rate all helpful posts.

[TyrosAdmin]

Both of you gave great information.

Will a priority queue starve the other queues if the bandwidth is not in use? 

IE 

Queue 1 using up 20% of overall bandwidth   (Priority Percent 40)

Queue 2 using up 5% of overall bandwidth     (Priority Percent 20)

Would there be 40% left over for everything else, or would the remaining bandwidth(75%) be free to be used?

[skarthic]

Yes. Thats the basic difference between a using "bandwidth" and "priority".

When configured for priority the bandwidth will be removed from the pool and will be used only for that class of traffic.

In your case the remaning 15% would only be used by other traffic.

Also if you use any Bandwidth command say bandwidth percent, it is calculated after deducting the bandwidth allocated

for priority traffic.

Here is an example.

policy-map Internet

match  class

priority bandwidth percent 50

match class

bandwidth percent 10

In this case if the interface bandwidth is configured as 1000.

So out of the 1000 Kb, 750 is usable

375K goes to class voice [even if not used'

Max 37.5K can be allocated for FTP but if not used the 37.5 can be used by other classes too.

HTH

Regards.