[nsp] rate-limiting

Steve Lim limmer at execpc.com
Fri Oct 10 10:54:48 EDT 2003


This link explains the Token Bucket concept:
http://www.cisco.com/en/US/products/sw/iosswrel/ps1828/products_configuration_guide_chapter09186a00800ca59f.html

This link takes you the Rate Limits specifically:
http://www.cisco.com/en/US/products/sw/iosswrel/ps1828/products_configuration_guide_chapter09186a00800ca59f.html#32499

Page down a couple of times for the Recomended Burst Values:
http://www.cisco.com/en/US/products/sw/iosswrel/ps1835/products_configuration_guide_chapter09186a00800bd8ed.html#1000977

Based on Cisco recomendation to avoid Tail Drops (and also approximate 
the configured rate), you would still configure as follows (of course 
you can tweak it if you want :)

interface fe0/0/0
rate-limit input 50000000 9375000 18750000 conform-action transmit
exceed-action drop
rate-limit output 50000000 9375000 18750000 conform-action transmit
exceed-action drop

This follows the following formula:

	normal burst = configured rate * (1 byte)/(8 bits) * 1.5 seconds
	extended burst = 2 * normal burst

Tweaking the replenish rate makes for some interesting times, especially 
   if you forget to tell the customer. :}

CAR or class map. Hrmm. How would this work out administratively, say if 
one where to apply said strategies if you grow to 20 customers on that 
router? Just absent-mindedly wondering.

Taa,

SL

Raymond, Steven wrote:

>>Hi Christopher,
>>
>>I know that when using traditional CAR, your policer 
>>bucketsize values 
>>are incorrect.  I assume the policer is implemented similarly 
>>with the 
>>modular QoS on the same platforms.
> 
> 
> Can anyone point to a CCO link that usefully describes the meaning and use
> of the bucketsize parameters?  In my trial & error testing with a smartbits,
> I found that the closest approximation to limit a FE circuit to say, 50Mb/s,
> would be something like this:
> 
> interface fe0/0/0
> rate-limit input 50000000 12500000 12500000 conform-action transmit
> exceed-action drop
> rate-limit output 50000000 12500000 12500000 conform-action transmit
> exceed-action drop
> 
> Where the 12500000s are the actual signalling rate of the media in terms of
> bytes/second.  With much smaller numbers, smartbits showed that achieved
> throughput was much lower than 50,000,000 bits/second at layer 2 (similar to
> the original poster's results).  These numbers usually let a little more
> than 50,000,000 bits/sec at layer 2 through, but I'd rather be over than
> under.
> 
> Furthermore, is there a difference in the configuration when doing straight
> interface rate limiting (CAR?) vs the modular QOS style?
> 
> Thanks
> 
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
> 
> 

-- 
<><><><><><><><><><><><><><><><><><><><>
Steve Lim - Network Engineer (Michigan)
Corecomm -An ATX Communications Company
On God's keyboard, he has a "Smite" button
-limmer



More information about the cisco-nsp mailing list