At 02:37 PM 12/12/2003, Fausto wrote:
>Why don't you all run EIGRP and redistribute OSPF (*) inside it. It is a bit
>more errorless prone and better to troubleshoot. At least it's my personal
>feeling.
Maybe because it's proprietary? Also, I've seen many networks on the verge
of collapse because someone thought EIGRP was best since it was
easy. EIGRP is a strategic sales tool for Cisco, and not much more than
that.
>(*) Just for VPN Concentrator and PIX devices.
>
>
>""Stephen Skinner"" wrote in message
>news:200312121658.hBCGwqT3020097@xxxxxxxxxxxxxxxxx
> > Ever thought about doing a e-learning(i hate that word) stream on
>Advanced
> > OSPF/BGP design...
> >
> > just a thought ...
> >
> >
> > >From: "Howard C. Berkowitz"
> > >Reply-To: "Howard C. Berkowitz"
> > >To: cisco@xxxxxxxxxxxxxx
> > >Subject: RE: Need help on OSPF cost! From miss OSPF [7:80374]
> > >Date: Mon, 8 Dec 2003 19:07:17 GMT
> > >
> > >At 6:43 PM +0000 12/8/03, Priscilla Oppenheimer wrote:
> > > >Ah, if only we had whiteboard capability on GroupStudy. I would love
to
> > >see
> > > >you discuss this with a whiteboard. It's a little hard to follow in
>text,
> > > >but I think I understand what you're saying.
> > > >
> > > >Howard C. Berkowitz wrote:
> > > >>
> > > >> At 12:29 AM +0000 12/8/03, Priscilla Oppenheimer wrote:
> > > >>
> > > >> >
> > > >> >Now regarding the network design, why do you have the routers
> > > >> in a circle?
> > > >>
> > > >> Actually, I find rings of moderate diameter to be good area
> > > >> 0.0.0.0
> > > >> designs. I've had good success, in high-availability networks,
> > > >> with
> > > >> putting a pair of ABRs at each site, linked by a high-speed
> > > >> LAN, and
> > > >> each with one or two WAN links to an adjacent site (different
> > > >> for the
> > > >> two routers). There are tradeoffs between load-sharing links to
> > > >> the
> > > >> adjacent router, for faster reconvergence if something fails,
> > > >> and
> > > >> linking each router to the "left' and "right" neighbors. This
> > > >> worked
> > > >> nicely in a seven-area domain.
> > > >
> > > >Hmm, a dual ring, with each ABR having a link to both an upstream and
> > > >downstream router (to use FDDI terminology)? Did you end up doing it
>that
> > > >way?
> > >
> > >Certainly in that manner, but with dual active rings (for want of a
> > >better term) doing per-packet load balancing for fast failover.
> > >Admittedly, that was in the days of having to process switch, but I
> > >_think_ it would still hold with CEF, or possibly using MLPPP.
> > >
> > > >I think it would have availability advantages because of the
diversity,
> > > >but convergence would probably be slower, if I can trust my gut
>instinct
> > >on
> > > >this.
> > >
> > >If the route is already in the table (e.g., per-packet process), it
> > >might not need to reconverge. Definitely if you were using L2
> > >multilink, the routing process would never see the failure, so OSPF
> > >reconvergence wouldn't have to run.
> > > > >
> > > >> If the number of links gets too high, or the hop distance too
> > > >> far,
> > > >> I'll put in (preferably) pairs of core routers and home the
> > > >> ABRs to
> > > >> them as well. If I do this and had dual links on the ring ABRs,
> > > >> I'll
> > > >> swing one link to the core backbone-only router.
> > > >
> > > >This could require a 3-dimensional drawing tool for clarification, but
>I
> > > >think I get it. :-) Would you just do one link to a core router or
>maybe
> > >2
> > > >(one to each core router?)
> > >
> > >Tradeoffs here include monetary cost of the links, the potential
> > >virtue of multilinking or convergence-friendly load sharing, whether
> > >or not the core router(s) were geographically dispersed, etc.
> > >
> > > >
> > > >With a network this complex would you even try to influence traffic
>flow
> > >be
> > > >messing with the costs, to bring us back to the original scenario
where
> > >the
> > > >costs were being messed with? I guess it would depend on requirements
>and
> > > >whether there's any need to maximize use of certain links.
> > >
> > >In general, I'd leave the costs alone other than possibly a true
> > >bandwidth derived metric. In one network, the core routers were DS-3
> > >(ultimately OC-3) while the ring links were (multiple) DS-1.
> > >
> > > >
> > > >Thanks,
> > > >
> > > >Priscilla
> > > >
> > > >> A core like
> > > >> that is
> > > >> also a good method of having a gateway to a single or limited
> > > >> number
> > > >> of firewalls. In days past, it also could be an ATM switch. I
> > > >> did
> > > >> this when the seven areas went to about ten, and there was a
> > > >> desire
> > > >> to centralize Internet access.
> > > >>
> > > >> >And if there's a good reason for doing that, why bother
> > > >> messing with the
> > > >> >costs? If the costs were all equal, the selected route would
> > > >> automatically
> > > >> >end up being the shortest. The metric would essentially become
> > > >> hop count,
> > > >> >and there may be nothing wrong with that. As others have said,
> > > >> and as you
> > > >> >know, you need to get to the bottom of why this network was
> > > >> designed the way
> > > >> >it was.
> > > >**Please support GroupStudy by purchasing from the GroupStudy Store:
> > > >http://shop.groupstudy.com
> > > >FAQ, list archives, and subscription info:
> > > >http://www.groupstudy.com/list/cisco.html
> > >**Please support GroupStudy by purchasing from the GroupStudy Store:
> > >http://shop.groupstudy.com
> > >FAQ, list archives, and subscription info:
> > >http://www.groupstudy.com/list/cisco.html
> > >
> >
> > _________________________________________________________________
> > Tired of 56k? Get a FREE BT Broadband connection
> > http://www.msn.co.uk/specials/btbroadband
> > **Please support GroupStudy by purchasing from the GroupStudy Store:
> > http://shop.groupstudy.com
> > FAQ, list archives, and subscription info:
>http://www.groupstudy.com/list/cisco.html
>**Please support GroupStudy by purchasing from the GroupStudy Store:
>http://shop.groupstudy.com
>FAQ, list archives, and subscription info:
>http://www.groupstudy.com/list/cisco.html
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=80785&t=80374
--------------------------------------------------
**Please support GroupStudy by purchasing from the GroupStudy Store:
http://shop.groupstudy.com
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
