04-28-2016 07:50 AM - edited 03-08-2019 05:32 AM
Hello community,
i have a problem i cannot understand actually. I often used limiting speeds of attached devices on catalyst switches directly on the port, worked properly on a 100MBit port like shown below, but on the 2960X i have to apply such a thing now, i seem not being able to use what i did so often before. The most obvious difference for me, i have GigaBit-interfaces.
The task is as follows:
I want to rate limit the gigabit port down to 200MBit for example, which i got working by playing around with the parameters and checking over and over with iperf. But, this is not an approach i can really use for the deployment in general, since i cannot test all the time until i figured out the correct settings. See especially my rate i policed the traffic ingressing that port...830000000 to achieve around 200MBit at the end...weird. As well i played around with the burst sizes, for example, didn´t really change dramatically the outcome. And even more strange, when i switch my policing rate to 860000000, for testing purposes...i finally ended up in having full gigabit speeds again.
Hmmmm...whats going on there? Can somebody explain me that, or tell me what i am doing wrong?
Is the 2960X not able to handle the policing properly, somehow interpreting my bitrates in a random way?
Is it maybe related to gigabit-speed on the IF, since exactly this worked perfectly in teh 100MBit world for me all the time?
I have to configure predictable results, because customer is planning to provide HIS customers customizable bandwidths on the switch ports directly, so maybe putting the ports to routed ports, or doing it on the upstream switch on the SVI is not what he wants actually.
And i absolutely cannot predict the results, keeping in mind i had to configure 830000000 for my target rate of 200MBit...makes not much sense for me.
Many thanks in advance, hopefully someone can shed some light on this...
Andreas
#####################################################################
My component:
WS-C2960X-48TD-L
c2960x-universalk9-mz.152-2.E3.bin
#####################################################################
Thats what i configured in my test, results as i wrote:
mls qos
!
ip access-list extended LIMITACL
permit ip any any
!
class-map match-all CLASS_TESTMAP
match access-group name LIMITACL
!
policy-map POLICY-LIMIT-200MBIT
class CLASS_TESTMAP
police 830000000 1000000 exceed-action drop
!
interface GigabitEthernet1/0/1
load-interval 30
srr-queue bandwidth limit 21
service-policy input POLICY-LIMIT-200MBIT
04-29-2016 01:06 AM
I would change the rate back to 200M, and adjust the burst amount up and down. You should come up with a ratio that works repeatedly for other bandwidth rates.
04-29-2016 02:53 AM
Hi Philip,
thanks for the answer, but see:
"As well i played around with the burst sizes, for example, didn´t really change dramatically the outcome."
I did exactly that, with no remarkable effects, besides of some MBit up and down...
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide