cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
1716
Views
5
Helpful
4
Replies

QOS-Real Life scenerio

fathar
Level 1
Level 1

Hi Guys,

We have got WAN consists of high speed serial and DS3 interfaces and I intend to limit multicast traffic max up to 30 percent of link.Rest reserved for other traffic. I am thinking to implement CAR to achieve this. Can some one please help to discuss this in detail like what should be the settings and where I should implement this policy, router end, L3 switch end etc.

Thanks for your help guys..

Cheers

4 Replies 4

mheusinger
Level 10
Level 10

Hello,

What you will implement is a policy in the DiffServ frame work. With DiffServ QoS you have to specify on each device a "rule set" on how to distribute the resources among the traffic.

In case you want to guarantee at least 70% of the interface bandwidth on a WAN line to non multicast traffic, then I would in fact not use CAR.

Example:

class-map match-all multicast

match ip address 101

class-map class-default

match any

policy-map OUT

class multicast

bandwidth percent 5

class class-default

bandwidth percent 70

fair-queue

interface Serial1/2/1

description WAN link

...

service-policy output OUT

access-list 101 permit ip any 224.0.0.0 15.255.255.255

What I thought is: what is wrong to give f.e. 80% of your bandwidth to multicast IN CASE there is no other tarffic? Probably nothing. So the above policy will give 70% of an interface to non multicast traffic and the REST to multicast IN CASE there is more traffic than the interface can handle. If there is no overload condition then service all traffic.

Apply this type of policy to any interface, where you like to influence ressource distribution.

Hope this helps! Please rate all posts

Martin

Hi Martin,

Thanks for the detailed response. I am little bit confused on few things and need to clarify..

(1) What is the advantage of using this technique

over CAR? Should I use POLICE command together to make it better?

(2) I am having my multicast server in one region and users Pcs in another, across the WAN. So where I should apply the policing and in which direction?Input or output

(3)should it be apply at L3 switch end or at router end as we need to make sure max. of 20%(for eg) traffic from each user end to this particular multicast stream.

Thanks

Hello,

1) the assumption I have - and which could be totally wrong - is, that it does not matter to you if multicast takes more than the specified 20% (or 30%) bandwidth - as long as you do not interfere with other traffic.

With a policer you restrict traffic to a certain rate - no matter what!

So let´s assume you have at a point in time ONLY multicast traffic - would you rather drop it at a max of 10 Mbps though your WAN link (assumed 45 Mbps) has 35 Mbps free capacity? This is what a policer would do. Tell me, when you want to achieve this.

What the configuration given does is so to say:

Multicast can take the ressources NOT USED by unicast traffic. So there could be 45 Mbps of multicast, but only if there is no unicast traffic. In case enough unicast traffic is there, multicast gets a maximum of 30% (bandwidth minus 70% reserved for all other traffic).

2) and 3) not sure you need a policer, in case you do, apply it outbound on the WAN links

Hope this helps. Please rate all posts.

Martin

P.S.: the above policy is a suggestion without intentionally dropping packets though resources are available (policer), but to restrict multicast traffic in case there is other traffic which needs protection.

In case you still are sure you need a policer use

policy-map OUT

class multicast

police percent 30

instead of the first example.

Thanks Martin. I will try both options.