[c-nsp] Rate limiting questions
justin at justinshore.com
Fri Oct 26 09:45:29 EDT 2007
Does anyone have any tricks for rate-limiting a pair of customer facing
L3 interfaces w/ HSRP across 2 chassis? I'm working on a co-lo setup
and I was thinking about how I'm going to rate-limit the customer and
still implement HSRP for access layer redundancy. Another engineer's
question got me wondering.
I was planning on applying the same rate-limits to the customer's L3
HSRP interfaces on both routers. However the customer could easily
point half their servers at the interface IP on the standby L3 interface
and it will gladly accept and route their traffic. Of course all return
traffic would come through the active HSRP interface but since most
co-lo traffic is upstream that rate-limit wouldn't prevent the
simultaneous use of both HSRP routers.
Is there a HSRP option to tell the standby router to only route traffic
when it's active? VRRP and GLBP would have the same problem I imagine.
Or is the solution to not pull the interface IPs out of the same block
as the floating standby IP, ie address the interfaces with RFC1918
addresses? This way the interface IPs wouldn't be routable from the
customer's server unless the customer bound the appropriate private
subnet to one of their interfaces and at that point their traffic
wouldn't be routable across the 'Net. I know this has been discussed
here dozens of times but I can't remember what the consensus was.
My other rate-limit question was about calculating the figures for CAR.
I once found a nice page that gave suggestions for what formula to use
in certain circumstances to calculate burst sizes. I can't find that
bookmark now and my Google-fu isn't turning up anything helpful. Does
anyone know the page or a similar one? Or does anyone have a suggestion
for burst size calculations?
More information about the cisco-nsp