Looking for real world OSPF example...

Have a question or want to start a discussion? Post it! No Registration Necessary.  Now with pictures!

Threaded View
Okay, I'm getting ready to test for the BSCI.  I understand much of the OSPF
stuff conceptually, but am having a hard time getting my hands around some
details without a large lab network to try it on.  I've set it up on a few
pieces of equipment with no issues, but it still doesn't give me the bigger
picture of a larger network.

What is the norm for larger networks that span multiple areas and are not
hub and spoke.  Do most people just use one large area 0 and no other areas?
One large area 0 and some other areas where they make sense?  Why?

In a larger area that spans multiple subnets, what is the best way to
identify which routers are the BDR and DR.  Is a BDR and DR elected once for
the whole area, or once for each broadcast segment?

Thanks for any input...


Re: Looking for real world OSPF example...
On Thu, 11 Jan 2007 04:32:08 +0000, Scooby wrote:

Quoted text here. Click to load it

A (show ip ospf interface) will show this info. DR and BDR are elected
from the router with the highest RID. The RID is the highest IP addresses
on any active interface, or the highest IP on any loopback interface(s).
So you can control DRs and BDRs with the IP addresses you assign to your
loopback interfaces.

Quoted text here. Click to load it

My understanding is that DR's and BDR's are formed from adjacences
(routers that can talk directly to one another), areas have nothing to do
with DRs and BDRs.

Re: Looking for real world OSPF example...

Quoted text here. Click to load it

    Using a "flat" design is not very usefull in a large enterprise network.
Some very large multinational companies use an IGP like OSPF as well     as
BGP to solve scalability issues.

Quoted text here. Click to load it
    More routes you have and more powerful your router has to be, so if you
have one flat area every router has the same LSA database ( rember that when
a change occur every router has to perform the Dijkstra algorithm ) and
every router must be "powerful"... "segmenting" using areas let you have
"big" routers only in the backbone.

Quoted text here. Click to load it
    Since the DR and the BDR has to make an adjacency with every other
router on the LAN segment it is preferred to use a powerful router if
there is a choice; you've also to consider what other rules the router has
to perform ( ABR, ASBR, BGP, ecc. ) to make the right choise... if     for
example you have a segment with 30 low-performing routers and 2 "high-end"
routers it's better to make those two router DR and BDR unless they are
overburdened with other heavy duties.

Quoted text here. Click to load it
    another way is to set the OSPF router-id or use interface OSPF priority
( eg. ip ospf priority <number> ) to force a router's interface to be
DR/BDR for a LAN segment.


Re: Looking for real world OSPF example...
Quoted text here. Click to load it

"my" valid reasons for split into different areas:
scale - 100 routers in any area (including the backbone) - practical limit
is less.
resilience - split highly resilient bits of the network into different areas
to low resilience
reliability - bits with likely flapping links and other reliability issues
should be in separate areas
adr aggregation - only places you can aggregate in OSPF are at ABRs or where
you import routes (eg blocks of statics for dialup).
loops - scale limits of an area reduce as you increase the number of "loops"
within it - because of the way LSAs circulate when a large topology change
Quoted text here. Click to load it

always try to minimise the size of the backbone - as it always tends to
grow, and it has to keep to the area size limits as well.

Quoted text here. Click to load it

having a neighbour in itself costs in terms of CPU, memory etc.

So run interfaces in passive where you dont need that subnet to carry
transit traffic (or use external for the interface instead).

Classic case is where there are dual central routers in a campus - you only
need a couple of the dozens of subnets to build a neighbour adj between the
2 switches. As a side effect this also controls which way traffic will go
during a fault.

Quoted text here. Click to load it
Not much in practice - because the election is sticky. so once something
gets elected, you only have another election when you force that device out
of being DR.

so all this does is alter the probability. if you tend to reboot your high
priority device a lot, it probabyl wont end up as DR.

if you have to control DR / BDR, set priority to 0 on some devices so they
cannot get elected.

Quoted text here. Click to load it

this means you can spread the DR around when the box has lots of
interfaces - but in practice this is just too much hassle to be worth it 99%
of the time.....
Quoted text here. Click to load it

stephen_hope@xyzworld.com - replace xyz with ntl

Site Timeline