cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
6999
Views
15
Helpful
11
Replies

How to configure QoS for VoIP

I'm doing some research on configuring QoS on our multi-site MPLS network and have a question. I think I've seen somewhere that our network provider's QoS settings would have to match our internal settings. So, if I'm trying to carve out 60% of a circuit for VoIP, our provider would have to do the same for that circuit, correct?

Thanks,

Chris

2 Accepted Solutions

Accepted Solutions

Mohamed Sobair
Level 7
Level 7

Hello,

what you are saying is absolutely correct. Now with MPLS QoS, there are three types:

1- Unfifrom Model.

2- Short PiPe Model.

3- PiPe Model.

The Service Provider choses the Type Of QoS to be used based on the Customer requirement and SLA Agreement.

Each on of those models has its own type and characterstics, but eventually to maintain your QoS marking across an IP/MPLS Backbone, the Service Provider needs to implement (Short PiPe Model) technique, which tunnels the QoS policy according to the CE (Customer Markings).

Regards,

Mohamed

View solution in original post

lgijssel
Level 9
Level 9

That's corect but if you need 60% for voip, you'd rather request an upgrade of the bandwidth.

Because voip needs the priority queue, other traffic is considerably throttled when you reserve 60% for priority traffic.

A more acceptable (safe) value would be 10% and you should really avoid going beyond 20%.

regards,

Leo

View solution in original post

11 Replies 11

Mohamed Sobair
Level 7
Level 7

Hello,

what you are saying is absolutely correct. Now with MPLS QoS, there are three types:

1- Unfifrom Model.

2- Short PiPe Model.

3- PiPe Model.

The Service Provider choses the Type Of QoS to be used based on the Customer requirement and SLA Agreement.

Each on of those models has its own type and characterstics, but eventually to maintain your QoS marking across an IP/MPLS Backbone, the Service Provider needs to implement (Short PiPe Model) technique, which tunnels the QoS policy according to the CE (Customer Markings).

Regards,

Mohamed

Okay, so we'll definitely need to contact our provider before we start implementing QoS. What would happen to traffic if we were to configure our end and the provider either didn't configure their end or misconfigured their end?

Thanks,

C

Christopher,

If this is to happen, then you will simply end up losing your markings, and you may also not having all your traffic priotrized as you want.

Regards,

Mohamed

lgijssel
Level 9
Level 9

That's corect but if you need 60% for voip, you'd rather request an upgrade of the bandwidth.

Because voip needs the priority queue, other traffic is considerably throttled when you reserve 60% for priority traffic.

A more acceptable (safe) value would be 10% and you should really avoid going beyond 20%.

regards,

Leo

Thanks both of you for your quick and thoughtful replies.

I think that the way it's been configured previously is through a policy map that set bandwidth to some number (I can look up the details if needed) that was a fraction of the total available bandwidth. Is there a 'best practices' kind of document out there? I'll show some of the config in a separate message.

C

here's some of the config....

policy-map Voice-Traffic

class Voice

  bandwidth 2500

class Business_Data

  bandwidth 3000

class class-default

  fair-queue

  random-detect dscp-based

policy-map QOS_INGRESS_LAN

class Business_Ingress

  set ip dscp af21

class Voice_Ingress

  set ip dscp ef                   

christopher_hall@earthlink.net wrote:

here's some of the config....

policy-map Voice-Traffic

class Voice

  bandwidth 2500

class Business_Data

  bandwidth 3000

class class-default

  fair-queue

  random-detect dscp-based

policy-map QOS_INGRESS_LAN

class Business_Ingress

  set ip dscp af21

class Voice_Ingress

  set ip dscp ef                   

You can probably do it like this as well:

class Voice

  bandwidth percent 10

etc...

The big advantage is that the traffic ratio's are preserved when changing the available bandwidth.

When using fixed numbers, you need to rewrite the entire policy.

With MPLS you probably also need shaping:

http://www.cisco.com/en/US/docs/ios/qos/configuration/guide/percnt_polcng_shapng_ps6350_TSD_Products_Configuration_Guide_Chapter.html#wp1025118

I will post a complete and working example shortly; its on another machine.

regards,

Leo

Here it is:

class-map match-any AF4x

match ip dscp af41

class-map match-any AF3x

match ip dscp af31

class-map match-any AF2x

match ip dscp af21

class-map match-any Voice

match ip dscp ef

class-map match-any Scavenger

match ip dscp cs1

class-map match-any AFclass

match ip dscp af21  af31  af41

!

!

policy-map AFsubclass

class AF4x

   police rate percent 20

     exceed-action set-dscp-transmit af42

class AF3x

   police rate percent 5

     exceed-action set-dscp-transmit af32

class AF2x

   police rate percent 5

     exceed-action set-dscp-transmit af22

policy-map STDqos

class Voice

  priority percent 10

class AFclass

  bandwidth percent 25

  service-policy AFsubclass

class Scavenger

   police rate percent 5

     exceed-action set-dscp-transmit default

     violate-action drop

class class-default

  bandwidth percent 50

  random-detect dscp-based

!

policy-map Shape10Mb

class class-default

  shape average 10000000

  service-policy STDqos

! class-map match-any AF4x
match ip dscp af41
class-map match-any AF3x
match ip dscp af31
class-map match-any AF2x
match ip dscp af21
class-map match-any Voice
match ip dscp ef
class-map match-any Scavenger
match ip dscp cs1
class-map match-any AFclass
match ip dscp af21  af31  af41
!
!
policy-map AFsubclass
class AF4x
   police rate percent 20
     exceed-action set-dscp-transmit af42
class AF3x
   police rate percent 5
     exceed-action set-dscp-transmit af32
class AF2x
   police rate percent 5
     exceed-action set-dscp-transmit af22
policy-map STDqos
class Voice
  priority percent 10
class AFclass
  bandwidth percent 25
  service-policy AFsubclass
class Scavenger
   police rate percent 5
     exceed-action set-dscp-transmit default
     violate-action drop
class class-default
  bandwidth percent 50
  random-detect dscp-based
!
policy-map Shape10Mb
class class-default
  shape average 10000000
  service-policy STDqos
!

This example uses a lot of features to play with. It does for example remarking and shaping.

Take out what you can use.

Leo

Very cool, Leo! Thanks a bunch!

Joseph W. Doherty
Hall of Fame
Hall of Fame

Disclaimer

The   Author of this posting offers the information contained within this   posting without consideration and with the reader's understanding that   there's no implied or expressed suitability or fitness for any purpose.   Information provided is for informational purposes only and should not   be construed as rendering professional advice of any kind. Usage of  this  posting's information is solely at reader's own risk.

Liability Disclaimer

In   no event shall Author be liable for any damages whatsoever (including,   without limitation, damages for loss of use, data or profit) arising  out  of the use or inability to use the posting's information even if  Author  has been advised of the possibility of such damage.

Posting

christopher_hall@earthlink.net wrote:

I'm doing some research on configuring QoS on our multi-site MPLS network and have a question. I think I've seen somewhere that our network provider's QoS settings would have to match our internal settings. So, if I'm trying to carve out 60% of a circuit for VoIP, our provider would have to do the same for that circuit, correct?

Thanks,

Chris

No, they don''t have to match, although most would configure them to match.

Since MPLS usually supports multi-site, it's possible traffic types might be different between sites or to some specific sites and traffic might also have asymmetrical flow rates.  For example, consider branch sites that back up their workstations to a HQ site.  Each branch, perhaps needs to set aside 20% of its QoS policy for this traffic's egress but perhaps the HQ site needs 10% of its QoS policy for this traffic's ingress.  In this example we're probably dealing with aggregation but other factors could be timezones, as all branches might not backup at the same time, and different bandwidth capacities for branches vs. HQ; both perhaps impacted necessary bandwidth reservations in the QoS policies.

Further, workstations restores are unlikely to use the same amount of bandwidth as the backups, so opposite direction QoS policy bandwidth reservations can be different too.

Thanks for the information, Joseph.

Review Cisco Networking for a $25 gift card