cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
4609
Views
0
Helpful
5
Replies

asr9000 subinterface rate limit application

Umit AYDINLI
Level 1
Level 1

Hi,

I have a lot of subinterface for customers. I want to use rate limit for subinterface. But I didnt see the rate limit command on asr9000 interfaces. Then I try to policy map so devices said 8 queue limit. How can do this?

2 Accepted Solutions

Accepted Solutions

Umit,

Can you explain more of what you are trying to do in the policies?  The issue will be, you are allowed 8 different queues on the -L or -TR cards.  But not every policy config would require a unique queue.  If you really do need unique queues for each sub-int then you would need the high Q cards. 

Here is a snippet from the comments in the link I referenced:

the Low Queue card has 8 queues per port. What consumes a queue on egress is the statements for priority, bandwidth and shape. For ingress, the priority statement will not consume a queue.

Policers and markers do not require a queue. Easy way to check it out is configure your desired pmap, apply it to the if and do the command show qos int , you get the hw programming and if there is a (unique) QID then a queue is consumed.

What 8 per port means is that on any given interface eg G0/1/0/1, Te0/4/0/0 or G0/1/0/9 each of those have 8 queues. Those 8 queues are SHARED between the subinterfaces! For instnace, I can have a pmap using 4 queues, then I can apply that to 2 subinterfaces on the same port before I run out.

Thanks,

Bryan

View solution in original post

the command rate-limit is not available in xR (and is being phased out in IOS also btw), the approach here for rate limiting is a policer config.

On the L cards you have 8k policers I believe and the B and E cards for Trident have about 32k/64k or something massive like that.

config is:

policy-map rate-limit100k

class class-default

police rate 100k

also consider adjusting the burst sizes to reasonable values, dont rely on defaults. the burst values depend on your RTT and the quality of service architecture document details out some suggestions on how to define the excess and committed burst values.

cheers!

xander

View solution in original post

5 Replies 5

Bryan Garland
Cisco Employee
Cisco Employee

Umit,

Yes, so you have a -L or -TR card.  These do have 8 queue limits.  Depending on what exactly you are trying to accomplish we can still do this.  I recommend reading over the following guide first. 

https://supportforums.cisco.com/docs/DOC-15592

Thanks,

Bryan

Hi Bryan,

I read before. But I didn't understand. how can apply to sub interfaces. I have subinterface about 1350.

If it possible could you provide a example configrations for the subinterfaces.

Umit,

Can you explain more of what you are trying to do in the policies?  The issue will be, you are allowed 8 different queues on the -L or -TR cards.  But not every policy config would require a unique queue.  If you really do need unique queues for each sub-int then you would need the high Q cards. 

Here is a snippet from the comments in the link I referenced:

the Low Queue card has 8 queues per port. What consumes a queue on egress is the statements for priority, bandwidth and shape. For ingress, the priority statement will not consume a queue.

Policers and markers do not require a queue. Easy way to check it out is configure your desired pmap, apply it to the if and do the command show qos int , you get the hw programming and if there is a (unique) QID then a queue is consumed.

What 8 per port means is that on any given interface eg G0/1/0/1, Te0/4/0/0 or G0/1/0/9 each of those have 8 queues. Those 8 queues are SHARED between the subinterfaces! For instnace, I can have a pmap using 4 queues, then I can apply that to 2 subinterfaces on the same port before I run out.

Thanks,

Bryan

Hi Bryan,

I'm using Te0/1/0/0.102 - Te0/1/0/0.1203 for subinterfaces and subinterfaces with metro ethernet customers. I want to need rate limit with metro ethernet customers.

This configrations runing perfectly cisco vxr series.

the command rate-limit is not available in xR (and is being phased out in IOS also btw), the approach here for rate limiting is a policer config.

On the L cards you have 8k policers I believe and the B and E cards for Trident have about 32k/64k or something massive like that.

config is:

policy-map rate-limit100k

class class-default

police rate 100k

also consider adjusting the burst sizes to reasonable values, dont rely on defaults. the burst values depend on your RTT and the quality of service architecture document details out some suggestions on how to define the excess and committed burst values.

cheers!

xander