I've been testing our different model of switches to allow for DSCP
transparency by using the "mls qos trust dscp" command on their
interfaces. All of the switches seem to support this properly and I can
tell when they're overwriting versus allowing the DSCP to continue
through but I came across one unique problem with the 3750 series.
I setup a traffic generator to send 95Mbps of traffic with DSCP EF (46)
across the different switches but when it hit the 3750, the egress
traffic was only ~4Mbps. Note that this is from a switch that has the
default configuration with mls qos turned on globally and the "mls qos
trust dscp" command put on the ingress and egress interfaces. Nothing else.
After reading up a bit, I found a command "srr-queue bandwidth shape"
that I could apply to the interfaces. After adding that command with
all 0s for the queues I was then able to receive all 95Mbps of traffic.
I noticed that the default values for that command are 25 0 0 0.
Is this something that I'm supposed to do if I just want to trust the
DSCP markings and not overwrite them on the 3750s? I don't have a need
for QoS (at the moment) and all my other Catalyst switches don't require
it so I'm just curious if I'm doing anything wrong for this model of
Catalyst. Note that I don't see this issue if I change my packets to
all go with DSCP of 0.
Thanks everyone.
Jose
_______________________________________________
cisco-nsp mailing list [email protected]
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/
