[ppml] Policy to help the little guys

[Stephen Sprunk]

Thus spake "Jo Rhett" <jrhett at svcolo.com>
> On Mar 20, 2008, at 12:21 AM, Ted Mittelstaedt wrote:
>> So, is this policy discussion actually solving a real problem, or are
>> we all doing someone's theoretical CCIE homework question for them?
>
> Not all big (or small) iron vendors do well with not using memory for
> refused routes.  One of those is the largest deployed vendor with no
> caps in their name.
>
> So your small guy may have a filter to drop anything longer than a /
> 24, but it's still eating up memory in his unit.  Unless the upstream
> is willing to filter it for him, *boom* goes his router one day.

There's several different types of memory.  The route will eat up BGP RIB 
memory, but it won't eat up FIB or main RIB memory.  Being short on FIB 
memory is generally the biggest problem, since fixing that generally 
requires totally new linecards or even routers; today, there's generally 
enough BGP RIB and main RIB memory, and if not it only takes replacing one 
or two cards per router -- not that one wants that expense either, but it's 
a lot lower than replacing everything.

S

Stephen Sprunk         "God does not play dice."  --Albert Einstein
CCIE #3723         "God is an inveterate gambler, and He throws the
K5SSS        dice at every possible opportunity." --Stephen Hawking