Agreed that there are few if any written materials that do a decent job of
explaining how OSPF works and why. Disclaimer - I have not read Tom Thomas'
book. But I have read a number of others, including the RFC. Plus my Lab
prep work.
One problem is that most of us don't understand the difference between
routing and routing protocols. Routing protocols are one means, but only
means, of getting routes into the routing table. Routing forwards or drops
packets based on the contents of the routing table.
Lack of understanding of how and why OSPF operates can lead to question like
"when there is a virtual link connecting a discontiguous area to the
backbone, does traffic travel across the transit area to the backbone, and
then back into the transit area?"
Area_2----area_1------area_0
If there is a virtual link from area 2 into the backbone, does traffic bound
for area 1 from area two have to go to area 0 first? The answer is two fold.
The area 2/area abr is now part of area 0. This allows area 2 to learn about
area 1 routes. And secondly, when a router receives a packet, it is the
routing process ( not the routing protocols ) that determine the forwarding.
Another question - do OSPF area numbers have to be unique? Answer is no they
don't. in a multiple area network, as long as there is an area 0, all other
areas can have the same number. there is nothing in an LSA that identifies a
route with an area. A corollary of this is that in a single area network,
there does not have to be an area 0. Why?
Got a good one. Do virtual links have to be point to point?
Area_3-----area_2----area_1------area_0
You can create a virtual link from area 3 to area 2 and a second link from
area 2 to area 0 - it works. Good example of this in Slattery's book.
Over time I have come to appreciate the Cert zone / Berkowitz approach,
which is to attain understanding of how things work. one might be surprised
at the number of people with very high level certification who can make, but
who don't know why it works.
Chuck
-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]] On Behalf Of John
Neiberger
Sent: Friday, June 22, 2001 7:32 AM
To: [EMAIL PROTECTED]
Subject: RE: OSPF Hub and Spoke [7:9268]
Yep, you are correct. I see now that as long as non-zero areas receive
LSAs from a router configured as area zero, then the routing tables will
be built and all should be well. This makes more sense to me, anyway,
but some of what I've read made the waters murkier rather than clearer.
How many times have you read "All interarea traffic must go through
area zero" ? I've read that in several places. It's almost true, but
not if you want to split hairs, especially since OSPF defines areas as
links, not routers. It should read that for loop-free routing to take
place, all non-zero areas must connect to area zero only. This is more
correct and doesn't imply that interarea traffic must cross an area zero
link.
Does that sounds about right? :-)
>>> "Chuck Larrieu" 6/21/01 11:15:31 PM >>>
John, this one's got me to thinking a little bit. Your kinda right but
kinda
wrong.
The areas are an OSPF structure, used for the building of the SPF
tables.
It's not that inter area traffic has to go through a discreet area 0,
but
that in OSPF in order for an area to learn about routes to another
area
there has to be an area 0 router in between them. It does not matter
if
there are a number of interfaces that are ABR's, or if there is a
discrete
and pure area 0.
With OSPF, all that matters is that the appropriate adjacencies are
formed,
and that the LSA's are processed and that the OSPF database is created.
If
all that occurs, OSPF routes will be placed into the routing tables. As
far
as the router itself is concerned, routing is independent of the
routing
protocols involved.
I've fooled with this in the past. I'll have to do another Q&D lab to
gather
some evidence, and post it here over the weekend.
In the meantime, for those interested in some in-depth discussion of
routing, Howard's white paper on Certification Zone is definitely
worth
reading. I have not seen the likes of it in any other source,
including
Doyle ( although it has been too long since I've read Doyle )
Chuck
-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]] On Behalf
Of John
Neiberger
Sent: Thursday, June 21, 2001 6:55 AM
To: [EMAIL PROTECTED]
Subject: Re: OSPF Hub and Spoke [7:9268]
Yes, I'm replying to myself.
While doing some reading it occurred to me why *not* extending area 0
across
the WAN links should not work. In OSPF, unlike IS-IS, an area is
defined by
links, not routers. The rule states that interarea traffic must go
through
area 0. Well, if areas are defined by links, then this means that
interarea
traffic must at least go across one link that is defined as an area 0
link.
In a hub-and-spoke environment with a single hub router, it seems to me
that
there just is no good way to use multiarea OSPF if you don't extend
area 0
across the WAN links.
At least, that's the way it appears at the moment.
John
| I'm having trouble wrapping my brain around a specific scenario and
I
| wanted to get your thoughts. Let's say we have a hub and spoke
network
| with a single router as the hub. There are five areas attached to
the
| backbone. It seems that we would have to extend area 0 across the
WAN
| links, but I'm wondering what would happen if we didn't.
|
| If we didn't, the backbone router would have no interfaces in area
0.
| I'm wondering if this would cause some major problems. I bet that
it
| would but I'm having a hard time thinking through what actual
problems
| might arise. Would this backbone router just "know" that it was area
0
| because it has interfaces in multiple non-zero areas and hence
behave
| correctly?
|
| One obvious problem is that the backbone router would be a member
of
| every area and would thus be pretty busy if the network got to be
very
| big. If we extended area 0 across the WAN link the backbone router
| would be protected from running SPF calculations everytime a remote
area
| had a link change.
|
| What other problems would arise? Would this even work at all? I
don't
| really have the tools to try it or I'd just attempt this chaos
myself.
| As you can guess, we run eigrp everywhere so I'm still clueless to
some
| of the workings of OSPF in a production environment.
|
| Regards,
| John
|
|
|
|
_______________________________________________________
Send a cool gift with your E-Card
http://www.bluemountain.com/giftcenter/
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=9527&t=9268
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]