gbonser at seven.com
Mon Oct 25 05:12:04 CEST 2010
> I'm not sure I agree with that. There is a fair bit that goes into the
> RIB entry - what routing protocol uses it, current state, and so on. I
> would expect the actual prefix/address to be a relatively small part
> that, perhaps 10-15%.
> In the FIB, you're approximately correct. The FIB contains the prefix
> (address and length) and a pointer to a list of zero or more next
Well, I was addressing FIB in my particular case. It is a vendor thing
in my case where the gear I am running can have 500,000 routes v4
(524,288 actually) routes (or 114,688 v6 routes) installed. The jump to
the next higher level gear is 1,000,000 v4 routes or roughly 1/4 that
many v6 routes. It can have a lot more routes (millions) from various
peers but that is the most it can have installed in the FIB. The cost
of that next jump is significant. And that 524,000 routes is for all
routes installed from all protocols for v4. That is BGP + OSPF + ISIS +
Now someone moving up to a Cisco 7600 with Sup 720-3BXL or from a
Brocade MLX to an XMR it wouldn't be a problem but that can be a pricy
jump. My point is that I believe we are going to see at first a lot of
resistance to routing table growth because it is going to be a problem
on dual stack routers with v4 fragging and v6 growing at the same time.
Now once v4 starts to shrink, then it will become less of an issue.
More information about the ipv6-ops