Can anyone point me to good documentation discussing setups for
large-scale POP's? We've got more than one class C's worth of PM2E-30's
now at our main POP, the site is still growing quickly, and I'd really
like to do this (some version of) right. I'd like to know more about
pros and cons for different ways of setting up addresses and routing.
I've dug around a decent amount, but haven't found anything complete
enough for my level yet. Direct opinions are very welcome--here are
things I'm wondering about:
Should we try to do it all for now on 2 contiguous class C's, with each
PM's assigned pool on the same class C as its ether0? We only have one
more contiguous to the primary ethernet class C, and will run out in the
not-too-distant future (if this way were overwhelmingly advantageous,
maybe we'd renumber in the future to a different set of four contiguous
subnets). This is the simplest way, & I could do it tomorrow,
supernetting the networks together with a 255.255.254 netmask. But it
won't last forever, and I know some people do it with assigned addresses
from a different class C, and I wonder what might be better about that.
Do some small security advantages kick in? I don't really know what's
involved in routing that setup, whether it can be done well with RIP or
is much better with OSPF, and what else would have to be in place on the
network routing-wise. (All of our PM2E-30's have just 1M of RAM
currently, not the 4M needed for OSPF, so that'd have to justify the
expense pretty well.)
And: we have a number of static-ip dialup clients, getting their ip from
a radius Framed-Address entry. Will any setup allow them to dial in and
route successfully even if they hit a PM with an ethernet address and/or
assigned pool of different class C from their address? And, well, I'm
just dumb about routing right now, but I haven't solved this with
"Framed-Route"--I don't really know if you can.
We have a setup that I don't consider so hot in place now: 5 of our 10
PM2E-30's are on a non-contiguous class C, routed onto their own local
ethernet, via a secondary interface on the Cisco that's the external
gateway for the whole joint. I don't like doubling all of their
ethernet traffic, and the static-ip customers have particular numbers in
the hunt group to dial, so they should hit a PM with the right class C,
but when the place fills up, that goes to hell.
Ok, I'm up too late and had better send this before it turns into a
novella, but I'd be glad to provide more detail if I've missed things.
Thanks for your time, any thoughts appreciated,
-- Ed Donovaned@capecod.net Cape Internet System Administration