Heh... ISP's shouldn't be runing an interior routing protcol over
their backbone, it'll snap or bite sooner or later. 8-)
Both OSPF and IS-IS are capable, whereas at the ISP level you're
going to find RIPv2, IGRP and EIGRP either broken, throw some
odd restriction or classful behavior in your face or make your
router worry too much about routing protocol overhead.
Between the two, maybe IS-IS has a little theoretical leg up, but
OSPF seems to be the one still seeing development intererest and
most of the nth-tier equipment vendors are doing RIPv2, BGP and
some OSPF subset real-soon-now. You probably don't want to get into
a situation where you need to run both. 8-)
One of the things you have to remember is that a lot of the stuff
in any (interior) routing protocol is designed to address a number
of scenarios, of which "ISP" is only one. Stuff like multiple
areas may have only limited real application for the ISP as opposed
to some coroporate "enterprise" model - it may work, but buy you
little gain for much pain.
>From my perspective (mostly linear regional network, multiple POP's
in each of several cites), the end evolution leads to:
a) RIPv2 (access servers or other clueless junk) is firewalled by
the customer aggregation router or something like an AS5300 with
two ethernets. Ideally only aggegate routes escape give or take
mobile static IP assignments.
b) OSPF (or ISIS) provide convenient automatic routing between the
POP's in a given city/suburban area. Clever regional IP block
assignments pave the way for aggregation.
c) BGP w/Confederations for each POP or city/suburban area provide
the internal routing between cities within your territory,
external routing and backup via other providers where indicated.
We're really not there yet (a) we have wretched /32's from the
access servers leaking all over our net, (b) historical address
assignments don't aggregate worth beans and (c) we still rely on
making our whole network on grand "area 0" for internal routing
vs. independent OSPF's for each city/suburban area and BGP glue.
One other thing we do that's slightly weird, is that when we
multi-home a customer, we will *only* talk BGP to them, not an
interior protocol, other ISP's will only to RIP or EIGRP or
(if they're brave) OSPF. The key here is that BGP is explict,
controlled and well documented/diagnosable - OSPF you might as
well shave your head, you're not going to have any hair left
by the time you figure it out.
Most of the OSPF documentation is illusory, BTW, there's a lot of
how-to and lets-explain-CIDR-again, but less in the way of clear
explanations, debugging hints, everybody knows and help for the
haunted.
George
> Date: Wed, 06 Sep 2000 22:57:59 -0400
> From: Steve Lilley <slilley@thrupoint.net>
> To: Cisco NSP List <cisco-nsp@puck.nether.net>
> Subject: [nsp] Advantages of IS-IS over OSPF
>
> Is there some advantage that IS-IS offers over OSPF for ISP
> backbones? I can't find anything documented...in fact, I can find
> VERY little documentation on IS-IS. This is quite a contrast to
> OSPF, where there is actually more information than I could ever
> want! :-)
>
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:16 EDT