Service Manual
Support for marking dot1p value in L3 Input Qos
Policy
In case the incoming packet is untagged and the packet which goes out to the peer is tagged, then the dot1p should be marked
appropriately using L3 Input Qos Policy. This is required because in the peer switch PFC will be generated based on the dot1p
value. Currently if the ingress is untagged and egress is tagged, then dot1p priority 0(default) will be added as part of the tag
header and from the next hop PFC will be based on that dot1p priority. Support is added to mark the dot1p value in the L3 Input
Qos Policy in this feature. Hence it is possible to mark both DSCP and Dot1p simultaneously in the L3 Input Qos Policy. You are
expected to mark the Dot1p priority when the ingress packets are untagged but go out to the peer as tagged
NOTE: L2 qos-policy behavior will be retained and would not be changed, that is we would not allow to set both DSCP and
Dot1p in the L2 Input Qos Policy.
Example case:
Consider that two switches A and B are connected back to back via a tagged interface. Consider the case where untagged
packets arrive on switch A, if you want to generate PFC for priority 2 for DSCP range 0-7, then you must need to match the
interested traffic using the class map. You should create an L3 Input Qos Policy and mark vlan dot1p as 2. You have to associate
both the L3 class map and L3 Input Qos Policy to queue 1 using the policy map.
In switch B, global dot1p honoring should be enabled, this will queue the packets on queue 1 as the dot1p will be 2 and PFC
should be enabled for priority 2. The policy map applied on switch A need not be enabled in switch B. When queue 1 in switch B
gets congested, PFC will be generated for priority 2 which will be honored in switch A.
You will not get the below CLI errors after adding this support:
DellEMC(conf)#qos-policy-input qos-input
DellEMC(conf-qos-policy-in)#set mac-dot1p 5
% Error: Dot1p marking is not allowed on L3 Input Qos Policy.
DellEMC(conf-qos-policy-in)#
You will also be able to mark both DSCP and Dot1p in the L3 Input Qos Policy:
DellEMC(conf)#qos-policy-input qos-input
DellEMC(conf-qos-policy-in)#set mac-dot1p 2
DellEMC(conf-qos-policy-in)#set ip-dscp 5
DellEMC(conf-qos-policy-in)#
Weighted Random Early Detection
Weighted random early detection (WRED) is a congestion avoidance mechanism that drops packets to prevent buffering
resources from being consumed.
The WRED congestion avoidance mechanism drops packets to prevent buffering resources from being consumed.
Traffic is a mixture of various kinds of packets. The rate at which some types of packets arrive might be greater than others. In
this case, the space on the buffer and traffic manager (BTM) (ingress or egress) can be consumed by only one or a few types
of traffic, leaving no space for other types. You can apply a WRED profile to a policy-map so that specified traffic can be
prevented from consuming too much of the BTM resources.
WRED uses a profile to specify minimum and maximum threshold values. The minimum threshold is the allotted buffer space for
specified traffic, for example, 1000KB on egress. If the 1000KB is consumed, packets are dropped randomly at an exponential
rate until the maximum threshold is reached (as shown in the following illustration); this procedure is the “early detection” part
of WRED. If the maximum threshold, for example, 2000KB, is reached, all incoming packets are dropped until the buffer space
consumes less than 2000KB of the specified traffic.
Quality of Service (QoS)
721