(1) CAR can only be done for the IPV4 packet, for it is not accurate for packets of other types.
(2) For the N1EGS2 board, to make CAR effective, the parameters CIR and PIR of the CAR must be equal. For the N2EGS2 board, the parameters CIR and PIR of the CAR can be unequal.
(3) The CAR is usually done at the INGRESS direction, at the EGRESS direction, it is not done. And normally, it is done at the IP side. It is of little significance to make CAR on the Vctrunk side.
See the following example:
In the previous figure, create private service between the IP interface of NEA and NE B. the binding bandwidth of the vctrunk is no less than 100M. If a 50M CAR at the ip1 interface on the NE A is created, the service from A to B is then limited within 50M, and from B to A, the service still can be 100M. If a CAR of 50M is made on the VCG1 on the NE A, then from SDH to VCG1 is the ingress direction of the CAR, and then from B to A, the service can only be 50M.
Normally, do not do CAR at the vctrunk. CAR means the traffic restriction on the network access side when the service accesses the network at the edge of the network. "vctrunk" belongs to the inside of the network, it is meaningless to restrict the traffic inside the network.
(1) CAR has a convergence period, it does not become effective immediately after being set, after several seconds, and it becomes stable.
(2) If the traffic is unstable and there is frequently burst traffic, the CAR is caused inaccurate, which is obvious to the N2EGS2 board.
(3) There is no such problem in interconnecting the meter, for the traffic on the meter is uniform. But the problem may occur in interconnecting with other equipment. Therefore, in interconnecting with other equipment, make clear the input traffic from the equipment interconnected. Do traffic shaping to the input data to reduce the burst traffic if you can.
(4) CAR is irrelevant to the service form, and it can be done for private service and layer 2 switching service. But it must be IP packet. CAR is also irrelevant to unicast and broadcast flow. For example ,broadcast flow also can be done as CAR( which is different from the COS, because the CAR is used to restrict the traffic in accessing the switching network inside the board, and the traffic restriction is not divided categorized in terms of the broadcast unicast packet in accessing. While the COS makes a priority allocation when outgoing the switching network. The broadcast is fixed as low priority, and the COS is ineffective to the broadcast packet).
(1) The board has two priorities of COS, 0�3 should be mapped as lower priority and 4�7 as higher priority.
(2) Two types of COS can be created on the board N2EFS0/N2EGS2 and the later EFGS board.
One is simple COS, it defines the priority of COS and flow binding. This type of COS has fixed priority after being bound. You can create many COSs of this type.
The other is COS of "vlanpri". Only one "vlanpri" can be created. In fact, it defines the priority mapping relation of the "vlanpri" domain. After binding this COS, you can transmit the service based on the "vlanpri" and the priority mapping relation.
(3) N2 board has multiple ID attributes and it can be used as N1 board. Note that when the N2 board works as N1 board, its COS cannot be used.
(4) Normally, the COS can be created at the IP interface, but it is meaningless to create the COS at the "vctrunk".
(5) In the service of layer 2, it is ineffective to create COS to the broadcast flow (including the flow that does not learn the MAC address and treated as broadcast). Only when the address is learnt and the broadcast changes to unicast flow, the COS is effective.
(6) For the lower priority of the COS, the NP discards packet. for the higher priority, PL215 discards the packet. (7) NP can discard the packet on average, but PL215 cannot. When the flows are all of higher priority, all the packet discarding is not on average.
See the following example:
There are following services: The IP1+VLAN1 and IP2+VLAN2 of end "A" use a same VCTRUNK to end "B". The bandwidth of VCTRUNK is only 100M, and the packet accessing IP1 and IP2 are both 100M. If IP1+VLAN1 and IP2+VLAN2 are both of lower priority, NP discards packet on average, at the end "B", you can see there are two flows of 50M. if IP1+VLAN1and IP2+VLAN2 are both of higher priority, NP does not discard packet and PL215 discards packet. Because PL215 discards packet randomly, seeing from end "B", the two flows are not averaged with 50M. The bandwidth of one flow may be more than 50M while the bandwidth of the other flow may be less than 50M.