QOS Marking clarification

[Pradeep H A]

Hi,

We have this configuration on our LAN interface:

Router#sh run int GigabitEthernet0/0

Building configuration...

Current configuration : 298 bytes

!

interface GigabitEthernet0/0

description "Customer LAN interface"

ip address 172.24.71.3 255.255.255.0

duplex auto

speed auto

service-policy output PolicyMap_0

end

Router#sh policy-map PolicyMap_0

  Policy Map PolicyMap_0

    Class ClassMap_0

      set dscp af31

    Class ClassMap_1

      set dscp af21

Router#

Router#show class-map ClassMap_0

Class Map match-all ClassMap_0 (id 9)

   Match   dscp af32 (28)

Router#sh class-map ClassMap_1

Class Map match-all ClassMap_1 (id 8)

   Match   dscp af22 (20)

As much as I know this would only remark all af32 traffic to af31 & af22 traffic to af21 that is entering LAN. Correct?

Can someone explain why would one do this in real time network?

By the way we are marking the WAN outbound conforming traffic to af31 & exceeding traffic to af32. Any link between them?

[allan.thomas]

Hi Pradeep,

The only rely sensible answer that I can make from the configuration is that you are remarking to ensure that these two traffic classes map to assured forwarding values that map to the low probability drop threshold. There are four assured forwarding classes each with three drop probabilities, high, medium and low.

For example AF32 and AF22 are assigned to the medium drop probability for the class AF3 and AF2, AF31 and AF21 are associated with the low drop probability, thus at the point of congestion high and medium are likely to suffer drops or policing compared to that of the low drop forwarding classes for each.

I can only assume that this is to ensure your transit network treat the traffic as appropriate such as the carrier if you have QoS applied over the WAN, and the carrier provides the sla that you have stipulated via the four different classes.

Regards
Allan.

Sent from Cisco Technical Support iPad App

[Pradeep H A]

Hi Allan,

Does that mean since WAN is of limited bandwidth, we are marking the traffic going out to WAN with different drop precedences & since LAN is usually 100 Mbps or more we no more need that differentiation & its unlikely that we need to drop anything in LAN?

Thanks in advance!

[allan.thomas · Accepted solution]

Hi Pradeep,

In most switch network environments you wouldn't ordinarily expect congestion, that said you will always where possible ensure that you utilise the hardware egress scheduling capabilities to move several different traffic classes through the LAN as appropriate and drop as appropriate, this is typically on egress due to multiple sources destined for the same destination.

As you said the WAN is of limited bandwidth thus you would mark or differentiate your traffic so that it maps to the appropriate queue and as agreed with your carrier. This is all dependent on whether the carrier trusts the ingress markings or remarks based on other criteria such as destination port? Either way you would always ensure that your sensitive or most critical data is classified and mapped to your low drop AF classes, and other traffic classes mapped to those which are acceptable to your organisation or qos policy where loss is acceptable to guarantee the more critical data.

Regards
Allan

Sent from Cisco Technical Support iPad App