CIDR,Sprint and the Big Guys.

Curtis Villamizar curtis at ans.net
Tue Apr 9 02:29:47 UTC 1996


In message <2.2.32.19960406011224.007175e4 at mail.cts.com>, "Kent W. England" wri
tes:
> At 08:25 AM 4/5/96 -0600, Tim Salo wrote:
> 
> >Perhaps, the [anticipated] consolidation of ISPs will be a significant
> >event in the efforts to control routing table size.
> >
> >-tjs
> >
> Do we need address leasing to hasten that process or should we let the
> market decide?  In any event, anyone buying up a bunch of ISPs will have a
> difficult time cleaning up the aggregation of addresses they inherit.
> 
> --Kent
> 
> BTW, I thought it was the updates, peerings and flaps that was the true
> problem, not the absolute number of routing table entries.


Not a problem at all in a well designed and implemented router.  With
say 30,000 routes and 100,000 updates in the queue, just empty the
inbound queue and decide what the end state is (including what get
dampenned and ignored now) and pass this end state on.  Install routes
at a lower priority.  Pass the end state on at a low priority.  Next
guy get far fewer update.  Dampenning actually puts a ceiling on it.

In a well designed router you can handle an arbirarily high amount of
routing change.  If it is really well designed you won't drop any
packets at all going to the destinations that are not changing.  The
only question is then whether you have enough memory to hold the
routing table if it continues to grow.

Today it is a problem.  :-)

Curtis



More information about the NANOG mailing list