Your comments help illustrate my points, answering the second point first
(because it is easier), if there is a need (or desire) to have multi-vendor
capabilities within your network then absolutely, you'll need to look at
something other than EIGRP (probably OSPF).
If you're getting SIA's then you don't have a healthy network because EIGRP
is not configured properly, plain and simple!
EIGRP will grow out-of-the-box only so big before it breaks - a point that
Cisco only recently started admitting.
If you simply do the following:
Router# config t
Router(config)# router eigrp 1
Router(config)# network 10.0.0.0
Router(config)# end
Router#
to say 100 or so routers in a redundant hub and spoke configuration - good
luck, I hope none of your wan links bounce.
I've seen more than one network like that fall to its knees as soon as an
instability occurred (7513's running at 99% CPU, cooking eggs on the RSP's
until you finally just powered them off). It is no different than OSPF
where you wouldn't want to have the same 100 routers all in area 0 (yes,
I've seen that too...)
Just as with OSPF, where you'd define different areas to break-up the
network, you'd do the same with distribute-lists and network summarization
(automatic or manual) in EIGRP.
So, it all comes back to the question from my original post:
>What is the reason for going to OSPF in this instance, stability problems
>with EIGRP or multi-vendor support?
If you haven't consumed your cup of Cisco kool-aid, haven't gotten the
religion of a Cisco end-to-end network then by all means, OSPF is probably
in your future.
If, on the other hand, EIGRP is melting down on a regular basis and the
CIO/CTO is seriously considering implementing RFC 1149 protocols then you're
faced with a decision. Migrate to OSPF where you're "forced" to implement a
better network design or fix EIGRP.
I spent a few years supporting a brokerage network that wasn't properly
designed; the initial designers - while well intentioned - didn't fully
understand IP addressing or the importance of bit-wise summarization. Long
story short, there were a little over 100 remote offices connected via
Frame-Relay to two hub locations, they were in the process of adding more
and were experiencing problems whenever a core T-1 would bounce. This is
about the time I was brought in to help with the routers. It all started
reaching critical mass (daily outages) at about the same time AT&T lost
their entire frame could for a day or so.
At the time, Cisco hadn't published the tricks to "proper" EIGRP network
design (or how to make a bad design still "work" with EIGRP), there was
still the attitude that EIGRP is "easy," simply configure it and forget it.
After multiple escalations within Cisco and getting to know the local
account team pretty well (seeing them daily) we were put onto Cisco's
"critical accounts" list. I don't know how many people on the list have
been in a similar situation - not a spot you really want to be in.
Basically, Cisco's customer support organization goes into overdrive,
because the customer is hemorrhaging $$ as a result of problems with the
Cisco equipment. We eventually met with the guys who wrote the Cisco Press
"Advanced IP Network Design" book (a group of some pretty "fart smuckers" by
the way) who explained what was going on and how to "fix" it.
Fortunately or unfortunately, you can "fix" EIGRP so that it works pretty
well even with a bad network design. We were given the "tools"
(distribute-lists, manual summarization, etc.) to fix the existing network
so that it was stable until the core problems could be addressed. I offered
proposal after proposal to correct the core problems (addressing,
distribution layer, etc) but they were never acted upon. Each time they
were shot down - layers 8, 9, and 10 in effect (Ego, Politics, and
Religion) - so that particular network is still in its "fixed" state and I
have moved on to other clients (after fighting for two years to get them to
address the real problems). The attitude of management was that they didn't
need to do anything, the network was working, which it was. I tried to
stress to them that it was only a matter of time before it stopped working
again and that we should work to correct the problems in a slow and steady
manner before being forced to implement a quick fix when it broke again, as
the consultant, I lost the battle.
So, I don't know which is better. If we had thrown out EIGRP in favor of
OSPF the network wouldn't have been able to run (for nearly three years now)
on the broken design, we would have had to split up the branches into areas
that were contiguously addressed and summarized into the core. Layers 8, 9,
and 10 wouldn't have been an issue because Layer 3 wouldn't have allowed it.
Given an all-Cisco network I would recommend EIGRP over OSPF any day. It
has faster convergence and more flexibility than OSPF - there have to be
some benefits to a proprietary protocol, right, otherwise no one would use
it.
I can't stress more that you have to have a solid network design to start
and you have to limit the query scope of DUAL (effectively breaking the
network into "areas") in order to have a stable, scalable network. If you
don't do these things, at some point it will break and you'll be looking to
migrate to OSPF.
Sorry for the long response...
I hope this helps,
Ben
-----Original Message-----
From: Carroll Kong [mailto:[EMAIL PROTECTED]]
Sent: Thursday, May 31, 2001 10:32 AM
To: R. Benjamin Kessler
Cc: [EMAIL PROTECTED]
Subject: RE: Migration EIGRP-OSPF [7:5724]
At 08:27 AM 5/31/01 -0400, R. Benjamin Kessler wrote:
>What is the reason for going to OSPF in this instance, stability problems
>with EIGRP or multi-vendor support?
>
>In my experience people seem to view EIGRP as "easier" than OSPF - while
>probably true in really small networks, networks these days just seem to be
>getting bigger and the same planning required for a successful OSPF
>implementation is required for EIGRP. I haven't seen too many companies
>with all-Cisco routers and a healthy EIGRP network looking to change
>things - thus the question above.
Well, a few points I would bring up is.
Stuck in Active problem of EIGRP. As the updates are being done, the
routers will stay in "active" mode (cannot receive new updates I
believe). If the EIGRP network is big, it must wait for the very last
router in the periphery to respond back. This could cause issues with
convergence time. You may have to modify the timers to increase the hold
time (which might cause bad convergence) since genuine requests might take
so long that they will get "zonked" out and the the router will delete it's
entry. This only happens in huge "AS" (in the EIGRP sense of an area of
sorts). So, if the idea of using OSPF and breaking into "areas" is bad,
you technically get the same issue with EIGRP, except in the form of ASes.
Also, you are running a proprietary protocol now. Although it seems to
work fine now. If say, they feel another vendor's product is superior in a
particular aspect of their network, they might be hard pressed or you will
need to do some redistribution/distribution lists which is probably going
to be difficult as well.
I suppose all in all it is still "easier" to use EIGRP. I agree
wholeheartedly with your statements. The cost of going to OSPF might seem
higher if they are really not that good with it. In that way it somewhat
validates them sticking to EIGRP.
-Carroll Kong
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=6747&t=5724
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
