From: arnaud.huret@xxxxxxxxxxxxxx
Date: Fri Aug 10 2001 - 10:56:01 GMT-3
Hi all,
The 50 routers max in an area limit is not to be considered now (if I
believe Cisco).
You can have an area 0 with 200 routers with no problem (and it does exist)
as long as your routers have enough memory and cpu power, which is the case
with 7xxx and 6xxx routers.
Of course if your area 0 is built on 2500 routers, you will surely
experience problems unless you have a highly stable network.
Generally, you put the strongest routers in the area 0, routers that can
handle many partial or full SPF recalculations.
You have to know that when a link is coming up and down in an area, the
full SFP calculation occurs only in that area, in other areas only a
partial recalculation will occur which is not so CPU-intensive.
In fact, the design has to based on what routers you are using and will
certainly vary if you want to perform summarization on ABR, if you have
distant geographical sites etc...
Arnaud
Internet
pvo@usermail.com@groupstudy.com - 10/08/2001 15:03
Veuillez ripondre ` pvo@usermail.com
Envoyi par : nobody@groupstudy.com
Pour : ccielab
cc :
ccc :
Objet : Re: OSPF Limits
What do you mean by a loopy discontiguous OSPF network? Can you provide an
example? Furthermore, you have to think of OSPF in terms of links and
their stability. Intra Area, OSPF does not deal with IP prefixes
specifically, but link states and link configurations. Certainly 100k Type
5's or Type 3's would present a huge burden on the network, however the
degree of burden s more related to the level of churn vs the sheer amount
of information. It really comes down to how often a router has to do full
or partial SPF's.
Pete
*********** REPLY SEPARATOR ***********
On 8/10/2001 at 12:02 AM Matt Wagner wrote:
>Obviously you have to have enough horsepower to run OSPF for a network of
>given size. But I'm pretty sure that if you had a bunch of high-end
>routers
>decked out enough to run BGP, they still couldn't handle 100,000 OSPF
>routes
>because the protocol wasn't made to handle a large, flat network of that
>scale. I am think that I read something to that effect in the Internet
>Routing Architecture book a while back, and I've heard it before, but I
>can't back it up with official documentation. I won't mind being wrong on
>this, either, since I'd rather see OSPF be more capable than less capable.
>
>In any case, the design of a network can have almost as much impact on the
>device trying to run the D algoryth as its resources. That is why a
>well-designed topology can have more routers in a given area that a loopy,
>discontiguous one. If you don't believe that, set up a test bed with a
>few
>hundred routers and let me know how it turns out.
>
>Matt
>
>
>
>
>
>A man said to the Universe, "Sir, I exist".
>The Universe replied, "The fact may be,
>but it inspires in me no sense of obligation."
>
>
>
>----Original Message Follows----
>From: jonatale@earthlink.net
>To: Matt Wagner <miguknom@hotmail.com>
>Subject: Re: OSPF Limits
>Date: Fri, 10 Aug 2001 00:02:10 -0700
>
>it is definitely a hoarse power issue (mem, BW, proc speed), the algorithm
>has
>no limits -- what would they be?? if you write good code (linked lists
>etc.)
>it
>can scale forever...
>
>Matt Wagner wrote:
>
> > Hey (group), let me know if I'm wrong about this, but the "50" rule is
a
> > generic rule. Whether you can support more or less than that depends
>on
>a
> > few things, like whether you are set up entirely hierarchically or
>whether
> > you are looping all over the place, the number of ASBRs, etc. I also
am
> > pretty sure that it has less to do with the horsepower of the routers
>you
> > are using and more to do with the protocol itself handling a database
>with a
> > certain degree of complexity well, which is the main reason you
>shouldn't
> > redistribute BGP into OSSPF.
> >
> > With proper design, you can probably handle more that 50 routers, and
>with
> > poor design, probably fewer. That's just what I have always thought,
> > though. Any comments would be appreciated.
> >
> > Matt
> >
> > A man said to the Universe, "Sir, I exist".
> > The Universe replied, "The fact may be,
> > but it inspires in me no sense of obligation."
> >
> > ----Original Message Follows----
> > From: "Dietmar Gaar" <strongbow71@gmx.at>
> > Reply-To: "Dietmar Gaar" <strongbow71@gmx.at>
> > To: <ccielab@groupstudy.com>
> > Subject: OSPF Limits
> > Date: Thu, 9 Aug 2001 11:59:12 +0200
> >
> > Hi,
> >
> > I4m playing around with OSPF last Days preparing for the Lab Exam.
> >
> > So in practice, when you have not a perfect but a good Design - how
>"big"
> > can an OSPF Interwork grow up ? Exist there some practical based Limits
>?
> >
> > I already know that there a many unknown Parameters to take a care of -
>but
> > maybe you can give me some scales...
> >
> > kind regards,
> >
> > Dietmar
> > **Please read:http://www.groupstudy.com/list/posting.html
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:31:48 GMT-3