Cisco Support Community
cancel
Showing results for 
Search instead for 
Did you mean: 
Announcements
Webcast-Catalyst9k
New Member

QoS 'nested' policies sanity check

Hi All,

I really want a sanity check as to if I ma going about my issue the right way.

At my HQ I have a MPLS connection to a SP. My remote sites have limited bandwidth. I want to create separate classes for each sites which average to certain speed. I then want inside each average to give voice priority always.

My test config is as follows.

class-map match-any voip

match ip dscp ef

class-map match-any qos_110

match access-group 111

class-map match-any qos_111

match access-group 112

policy-map child_111

class voip

priority 48

class class-default

fair-queue

policy-map child_110

class voip

priority 48

class class-default

fair-queue

policy-map parent

class qos_110

shape average 128000

service-policy child_110

class qos_111

shape average 128000

service-policy child_111

The way I understand it, is that I have average the BW to each site to 128k, and within that BW when there is voice traffic it always gets priority.

As the 'rest' of the traffic, I'm not that concerned, hence the 'class-default' class.

3 REPLIES
Silver

Re: QoS 'nested' policies sanity check

Hi,

It looks a bit overly complicated to me,

Basically you want to create a top level policy-map with class class-default, and shape to desired rate.

Secondly configure a class-map and match your voip traffic within, then create a new policy-map and configure LLQ as required, ie, priority 48.

For example

class-map cvoip

match ip dscp ef

policy-map pvoip

class cvoip

priority 48

policy-map ptop

class class-default

shape average 128000

service-policy pvoip

then attach ptop to the interface outgoing...

The logic here is that when and only when the top level policy-map is shaping, ie you are sending more the 128k pps, then it calls the 2nd level polic-map and traffic within the cvoip class gets priority, everthing else gets best effort.

HTH

LR

New Member

Re: QoS 'nested' policies sanity check

Thanks for that.

However, how do I apply it for several sites based on an ACL, so that each sites gets the same policy (voip 48 priority and shaping at 128)?

Regards

Super Bronze

Re: QoS 'nested' policies sanity check

If your MPLS provider supports QoS, your policy should be configured to work in conjunction with theirs. (Without any MPLS QoS, a correct QoS policy is difficult to support if you can have remote site to remote site traffic.)

Whether the provider supports QoS or not, you'll still need to address interface congestion on your router.

You'll want to insure your VoIP goes first both at your interace and upon MPLS egress. (Again, the last is difficult to control outside of working with the provider's egress QoS.)

Without MPLS egress QoS and assuming there's no traffic between remote sites, you'll want something like:

(NB: syntax might be incorrect)

class-map match-any qos_110

match access-group 111

class-map match-any qos_111

match access-group 112

class-map match-all voip_110

match ip dscp ef

match class qos_110

class-map match-all voip_111

match ip dscp ef

match class qos_111

policy-map HQ

class voip_110

priority 48

class voip_111

priority 48

class qos_110

shape average 80000

class qos_111

shape average 80000

(if you don't oversubscribe the local interface, this configuration could be used)

class-map match-any voip

match ip dscp ef

class-map match-any qos_110

match access-group 111

class-map match-any qos_111

match access-group 112

policy-map child

class voip

priority 48

class class-default

fair-queue

policy-map parent

class qos_110

shape average 128000

service-policy child

class qos_111

shape average 128000

service-policy child

146
Views
0
Helpful
3
Replies
CreatePlease to create content