How To Use QOS Counting To Verify Layer 3 Packets Sent And Received

By esupport posted Mar 30, 2020 02:28 PM

  
Requirement:

Any Comware 5 or Comware 7 platform switches

Issue

How and where QoS has to be applied to influence queue selection/prioritization of traffic?

What are the rules and best practices to apply the QoS policies for traffic prioritization locally and over the path being traversed?

Cause

To place the desired type of traffic into queues for its better utilization and most importantly prioritization.

To verify the desired traffic is being prioritized on the correct device and is channeled across the correct queue.

 



Solution:

On the Comware 5/7 platforms, traffic can be prioritized and/or remarked using the below :

  • DSCP
  • DOT1P
  • LOCAL PRECEDENCE

 

Please NOTE: Prioritization and remarking are NOT the same.  Prioritization implies the packets are to be moving in the desired queues but remarking is simply changing the DSCP values of the packet only to be prioritized on the next hop’s egress.

Remarking will not change the priority internally on the switch of anything that comes in ingress unless local-precedence is defined.

Local precedence remarking influences prioritization locally on the switch and so does dot1p remarking.

QoS trust dscp is enabled on a port, and has a packet that comes in with a dscp value XX, will prioritize according to dscp value XX. If there is a remark on the inbound policy, the packet will be remarked with the new dscp value to be prioritized with the remarked value on the next-hop.

A packet is also prioritized if it is a ‘tagged packet with a COS priority’

 

The queue handling behavior can be summed up with the below rule-set :

  • Inbound policies should be used either at port interface or vlan interface
  • When inbound policies are used to remark traffic (port, vlan, global), local precedence needs to be remarked in order for prioritization to occur within the switch.
  • If a packet arrives w/o any  dscp or COS priority, it will take the best effort queue
  • If a packet arrives in a port that has a dscp value pre-defined and qos trust dscp is not on the port, it will not be prioritized within the switch.
  • If a packet arrives in a port that does not have a dscp value pre-defined,  it will not be prioritized internally in the switch, even if the inbound policy includes a dscp remark.
  • Qos trust dscp, must be applied to the interfaces, as packets that are remarked will then be trusted on the next-hop switch
  • Qos trust dscp must be applied to all ports where voice traffic enters, as the EF value is predefined on the packet  when phone is booted.
  • Qos trust dscp is enabled on a port, and has a packet that comes in with a dscp value XX, will prioritize according to dscp value XX.
  • Qos trust dscp is enabled on a port, and has a packet that comes in with a dscp value XX, will prioritize according to dscp value XX, if there is a remark on the inbound policy,  the packet will be remarked with the new dscp value.
  • If a packet arrives with a cos value, it will retain the priority within the switch
  • Qos remarking to a dot1p priority does not need to be remarked w/local precedence for it to be processed internally.  Testing did not matter whether qos trust was on or not inbound.

 



Configuration:

Test setup and results :

 

Facts :

  • Layer two topology passing VLAN 10 across and their subnet being 10.10.10.0/24
  • Continuous PING from 10.10.10.2 to 10.10.10.4 is in place.
  • 10.10.10.4 is a PC issuing PING with no pre-defined DSCP value.
  • Local precedence is expected to prioritize the ICMP traffic onto queue 5 due to the command “local-precedence 5”
  • Remark dscp af43 is supposed to prioritize the same ICMP traffic onto queue 4 on the next hops’s egress ports.

 



Verification

Test setup and results :

Facts :

  • Layer two topology passing VLAN 10 across and their subnet being 10.10.10.0/24
  • Continuous PING from 10.10.10.2 to 10.10.10.4 is in place.
  • 10.10.10.4 is a PC issuing PING with no pre-defined DSCP value.
  • Local precedence is expected to prioritize the ICMP traffic onto queue 5 due to the command “local-precedence 5”
  • Remark dscp af43 is supposed to prioritize the same ICMP traffic onto queue 4 on the next hops’s egress ports.

 

Notice, no packets are forwarded over queue 5 before applying the local-precedence 5 :

 

 

 

Upon applying the QOS policy with behavior local-precedence 5, the classified ICMP traffic get re-directed to Queue 5 as you’d need it be :

 

Now, since there was no pre-defined DSCP value from the PC 10.10.10.2, the remark will be applied on the switch but carried to the nexthop and would be in effect only on its egress ports.

To prove this I have applied the DSCP value of AF43 so the nexthop switch chooses queue 4 specifically.

 

Notice the nexthop switch below with 0 packets in queue 4 before applying the “remark dscp af43” on the traffic behavior :

 

 

 

 

 

Upon applying the “remark dscp af43”, we see packets being prioritized on queue 4 successfully.

 

 

0 comments
0 views