To classify IP packets, you configure the ingress label switching router. Packets are received at the ingress router as IP packets and sent as MPLS packets. To perform the configuration, use either of the following features:
â¢Modular QoS CLI, the newer and more flexible method-Use this method if you do not want to consider the rate of the packets that PE1 receives.
â¢CAR-Use if you want to consider the rate of the incoming packets:
-If a packet conforms to the SLA between the service provider and the customer (that is, the packet is in-rate), the service provider gives the packet preferential treatment when the network of a service provider is congested.
-If a packet does not conform (that is, it is out-of-rate) and the network is congested, the service provider might discard the packet or give it less preferential treatment.
You cannot do this, as policing by its very nature does not queue packets that are deemed to exceed the limit, so there would be no queues to take the packets from. Whereas shaping does queue the excess packets so you can do LLQ/CBWFQ here.
However in your case, I think you have got confused on where to do policing and where to do LLQ/CBWF.
In the example network
from traffic from CE1 to CE2 you would implement policing on ingress from CE1 to PE1, you would do LLQ/CBWFQ on egress from PE1 to P1, you would then do LLQ/CBWFQ on egress from P2 to PE2, and again on egress at PE2 to CE2.
from traffic from CE2 to CE1 you would do it in the reverse direction.
You need to think of the traffic flow as being bi-directional.
Thank you for your detailed reply. I totally agree with your point of view. I have following additional questions:
1. Ingress on PE1 from CE1, policing would be implemented as well as marking of MPLS EXP would also required. May I use rate-limit for policing and MQC for marking. Or I have to use MQC for policing and marking both. This is what the original confusion is?
2. From PE1 - P1 - P2 - PE2 - CE2, LLQ/CBWFQ would be used based on EXP or IP Prec or it could be mix?
1. you can use either rate-limit or MQC, they both do the same job. Rate-limit is the older method of the two, it would make more sense to use the MQC as this is the way forward for QOS config.
When you configure the policing via MQC you specify the actions to take on the packet based on whether it conforms, exceeds or violates the traffic (again you can do this via CAR or MQC), you would set the MPLS EXP using the imposition, which would mean that the EXP you specify would be set on the top (IGP) label, and the bottom (VPN) label.
2. From PE1-P1-P2-PE2 you would do LLQ/CBWFQ based on the MPLS EXP, as this is the only field you can classify on, since the IP packet in not inspected whilst in transit, ie all forwarding of the packet is based on the MPLS label.
For PE2-CE2 there are 3 or 4 different ways that you could do this, ie you can do LLQ/CBWFQ based on the ingress (P2-PE2) MPLS EXP or you can do the LLQ/CBWFQ based on the original IP dscp/ip precedence marking on the IP packet.
Thank you for your email. Query 2 is clear now. As far as query 1 is concerned, I am still confused that how policing and marking of EXP can be done simultaneously using ingress policy-map on PE1 from CE1.
Let suppose I want to police the incoming traffic on PE1 from CE1 to 1MB and at the same time mark the voice traffic with EXP value 5 and Network Management traffic with EXP value 4 and Oracle traffic with EXP value 3 while rest with EXP value 0.
Introduction: The "external-out enable" command is available for
configuration under the "router ospf process" in case of the IOS-XR
operating system. This command basically enables advertisement of
intra-area routes on the device as external routes in th...
Introduction Basic configuration for netflow Scale parameters for
netflow Netflow support Architecture Packet flow for netflow Inside the
LC CPU Netflow Cache size, maintenance and memory Sample usage Cache
Size Aging Permanent cache Characteristics Which...