[c-nsp] Understanding burst numbers in rate-limit

[Steve Bertrand]

Hi all,

This is a rather basic question, but it's my first attempt at rate
limiting a non-receive interface on a Cisco device.

Cisco recommends a particular mathematical formula when using the
rate-limit command, which I've executed in the example below.

After scouring the web and Cisco docs, I honestly can't make heads or
tails of what the 'burst normal' and 'burst max' actually mean. I've
tried to apply the numbers against time, but I just don't get it.

Can someone provide a decisive mathematical method as to how these
numbers are used, and when/why?

I have goals outside of the scope of this question, but understanding
this will allow me to eventually apply policy to it, with the ability to
answer questions and tweak as necessary.

The following example is based on a 100kb/s rate-limit:

# sh int fa0/1.767 rate-limit

FastEthernet0/1.767 xxx xxx - VLAN 767
  Input
    matches: all traffic
      params:  96000 bps, 18750 limit, 37500 extended limit
      conformed 103433 packets, 47968523 bytes; action: transmit
      exceeded 22507 packets, 12933277 bytes; action: drop
      last packet: 4876ms ago, current burst: 7012 bytes
      last cleared 05:57:23 ago, conformed 17000 bps, exceeded 4000 bps
  Output
    matches: all traffic
      params:  96000 bps, 18750 limit, 37500 extended limit
      conformed 49327 packets, 2927605 bytes; action: transmit
      exceeded 0 packets, 0 bytes; action: drop
      last packet: 4392ms ago, current burst: 0 bytes
      last cleared 05:21:18 ago, conformed 1000 bps, exceeded 0 bps

Cheers,

Steve