From: Brian Hescock (bhescock@xxxxxxxxx)
Date: Sun Aug 05 2001 - 09:30:47 GMT-3
Assuming you have enough memory, etc, and a proper design you should be
able to do more than 40 neighbors. I've seen people use 200 eigrp
neighbors on a router in a point-to-point environment. The key thing is a
good design. Run auto-summary if at all possible. That means a good ip
address plan. If you can't, be sure to manually summarize wherever
possible. That's the single most significant mistake people do, they turn
off auto-summary but don't replace it with manual summarization. If
you're using private ip addressing there really shouldn't be a reason why
you can't use auto-summary, i.e. you shouldn't end up with a discontiguous
network. Also, don't run eigrp on very slow-speed serial links, such as
32k links nor on 64k links that are saturated. Other factors include
dual-homed remote sites. Summarize down to dual-homed resites (as you
should anyway) and use distribtute lists so the remote site only advertises
networks attached to / behind it and not be a possible transit area. By
that I mean don't let the remote site advertise routes that come in from
the core back out the other link back to the core and vice-versa. EIGRP
has to keep track of every possible path and during SIA events it's a huge
problem with dual-home remotes otherwise. Take a look at EIGRP stub as
well, good feature. I can't recall off the top of my head if it came out
in 12.0T or 12.1T but it's worth looking up. Here's a url that discusses
it:
http://www.cisco.com/univercd/cc/td/doc/product/software/ios122/122cgcr/fipr_c/
ipcprt2/1cfeigrp.htm#xtocid2271320
Also, run the latest version of IOS. At around 12.1(6) or so we handle
EIGRP SIA much better. Previously if we had a remote router that couldn't
respond to a query we would end up killing every neighbor in the patch back
up to the router that first originated the query. The reason for that is
if you don't receive 100% of the responses to the queries you send out you
can't send your own response upstream and so on. In the later code we only
kill of the offending router at the far end that couldn't respond for
whatever reason (saturated link, high cpu, low memory, etc). Ideally you
would want to upgrade all your routers but it is backwards compatible so
you don't have to (but you won't get the full benefit, naturally).
Brian
Annu wrote:
> Ryan,
>
> Ur question has the answer..there is no definitive
> number bcoz it depends on multiple factors.
>
> At one of our sites we have 48 EIGRP neighbours on
> couple of routers and with latest code of IOS and we
> run into problems sometimes (of course not all
> adresses are summarized for various reasons). I
> remember talking to one of Cisco consulting enggs
> (RTP) and they had mentioned 40 is good #. As it
> depends on factors like memory,ur router model,
> traffic and maybe even network stability definitve
> numbers are not there i guess.
>
> Sombeody correct me if i am wrong.If u get any info
> let us know.
>
> bye,
> Annu
>
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:31:45 GMT-3