cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
1863
Views
0
Helpful
7
Replies

wan qos classification and marking

thomas.fayet
Level 1
Level 1

hello all.

I have a question regarding qos over frame-relay .

Let's say we dont use the same classification and marking on both sides of the network.

On side A you put traffic X in REAL-TIME class (dscp=ef)  and on side B you put traffic X in DATA class (dscp af41) , so the marking and classification is not the same.....

What will happens in case of congestion of the frame-relay link ?

Second question , is the qos mechanism on the cisco router ALWAYS active or is it just active when you have congestion on the link ?

thanks

2 Accepted Solutions

Accepted Solutions

jkillion
Level 1
Level 1

Hey Thomas -

1)  Your QOS congestion management or avoidance mechanism is completely independant of your global marking scheme.  Hence, router A's congestion management (for example) will treat packets in accordance with their markings in relation to what you have specified for the policy to do - ie. guarantee bandwidth, police etc.  If you change markings before packets reach router B, router B could care less.  It is still going to simply follow whatever QOS policy you have configured.  If you tell it to guarantee 10Mb for DSCP EF, it makes no difference if EF was classified VoIP at previous points in the network, but now is being utilized for FTP marking.

Quite simply - the router will do what you tell it to do at each point in the network.

2)  "Second question , is the qos mechanism on the cisco router ALWAYS active  or is it just active when you have congestion on the link ?".....Depends.  Congestion avoidance (such as policing / shaping) will occur irrespective of how utilized your interface is.  Congestion management (such as CBWFQ) only kicks in during times of congestion.  Then theres WRED, which is also congestion avoidance....WRED sits somewhere between policing and CBWFQ in that it only starts at a predefined threshold, and ramps up in force as congestion increases.

HTH

View solution in original post

Hi,

This is result of poor design. One thing you have keep in your mind when designing QOS is desing end to end QOS solution, not the one you are talking about. No administrator would like to have same category of traffic marked differently within their network. They would also ensure that no one sending wrongly marked packets when they are connected to other networks. QOS design should be neatly defined in terms of boundary and consistent markings for traffic types.

HTH

Hitesh Vinzoda

Pls rate useful posts.

View solution in original post

7 Replies 7

Mohamed Sobair
Level 7
Level 7

Thomas,

whenever a link on the frame Relay is experiencing congestion, the Frame relay Switch set the becn and fecn bit based on the type of the traffic and it direction.

1) If your router is to respond to those becn and fecn Bit in th frame relay header, then you router will drop back to the actual Cir value provided by the Service provider, however, if Not then , your excess traffic is subject to be dropped.

2) The QoS is active when you have a congestion.

HTH

Mohamed

thanks but this is not replying to my question .

what will happens to the same type of traffic beeing classify differently on both end of the netowrk .

let say I put sip signalling in real time class on side A and on class data1 on side B (different claasification and marking)

what happens in case of congestion on one end of the network ?

let me know if that is not clear .

rgds

On the end that experiencing congestion, The router will be subject to the policy you have set for the QoS ONLY on the end experiencing congestion.

for example, if your real time application have QoS policy set, then it will be syubject to that policy whenever a congestion occurs.

Please ask if you have any other question,

HTH

MOhamed

jkillion
Level 1
Level 1

Hey Thomas -

1)  Your QOS congestion management or avoidance mechanism is completely independant of your global marking scheme.  Hence, router A's congestion management (for example) will treat packets in accordance with their markings in relation to what you have specified for the policy to do - ie. guarantee bandwidth, police etc.  If you change markings before packets reach router B, router B could care less.  It is still going to simply follow whatever QOS policy you have configured.  If you tell it to guarantee 10Mb for DSCP EF, it makes no difference if EF was classified VoIP at previous points in the network, but now is being utilized for FTP marking.

Quite simply - the router will do what you tell it to do at each point in the network.

2)  "Second question , is the qos mechanism on the cisco router ALWAYS active  or is it just active when you have congestion on the link ?".....Depends.  Congestion avoidance (such as policing / shaping) will occur irrespective of how utilized your interface is.  Congestion management (such as CBWFQ) only kicks in during times of congestion.  Then theres WRED, which is also congestion avoidance....WRED sits somewhere between policing and CBWFQ in that it only starts at a predefined threshold, and ramps up in force as congestion increases.

HTH

Thanks for the replies .

Ok so now let s say both routers on both ends on your network are using the exact same qos configuration (rt and 3 data classes D1 , D2 , D3 with same bandwidth repartition) but router A is classifying the SIP protocol (for example) in RT and router B is classifying the same SIP protocol in D1 .


Let's also say that the link in ROUTER A is congested and the RT class if full (provisoned for 2 voice channels and both channels are "occupied")  and a sip discussion occurs between router A and router B with router A as source.


Router A uses the following configuration for it is RT class


Policy-map Queuing

class VOICE-OUT

priority

police 24000 3000 conform-action set-dscp-transmit 46 exceed-action set-dscp-transmit 46

exit

!

class-map match-any VOICE-OUT

match ip dscp 46

I though that when one of the link was congested the discussion has to be done on the same class .......so NOT beeing consistent in the marking and classifying scheme on both end of the network was not the bext thing to do ....what will happens to a paquet beeing marked EF on router A , getting to router B and now Router B replied (still using SIP) but with a packet beeing marked as D1 ........


last question , let s also say in router A the RT class is full but not D1 D2 and D3 , can the RT class use the bandwidth assigned to the data classes ?

Is the policer avoiding the RT class to do so ?


thanks


no one ?

Hi,

This is result of poor design. One thing you have keep in your mind when designing QOS is desing end to end QOS solution, not the one you are talking about. No administrator would like to have same category of traffic marked differently within their network. They would also ensure that no one sending wrongly marked packets when they are connected to other networks. QOS design should be neatly defined in terms of boundary and consistent markings for traffic types.

HTH

Hitesh Vinzoda

Pls rate useful posts.

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: