[c-nsp] OSPF scalability

[Tantsura, Jeff]

Hi,

Unless you are doing TE, hierarchical OSPF is GOOD.
The only way to filter in link state protocols is by using different areas
The only way to summarise in link state protocols is by using different
areas.

--
Jeff Tantsura  CCIE# 11416
Senior IP Network Engineer


-----Original Message-----
From: Andris Zarins [mailto:andris.zarins at microlink.lv] 
Sent: 06 September 2005 13:19
To: Kristofer Sigurdsson
Cc: cisco-nsp at puck.nether.net
Subject: RE: [c-nsp] OSPF scalability

Hi,

OSPF works finw with several thousand routes, that's not a problem. 

One thing that I definately recommend is NOT to split network in several
OSPF areas. There are no winnings if you do that. None at all. Only
thing you get is more complex configuration and more chances to run into
some trouble. Areas would be recommended if you have 250+ routers, just
to scale network and maybe ease administration, or for example if you
have some stub segments. If all areas should have all routes, and there
are only some routers in each of them - stay with one area.  

Also I wouldn't recomment iBGP as IGP, just because it's a bit slow.
Yea, of course it gives you more granular control of routing information
and lots of other nice features, and of course you can tune timers to
make it faster, but I'd stay with OSPF as IGP, especially - if user
sessions are dynamic and routing info changes almose every second ;) .. 

Hope this helps, 
Andris






-----Original Message-----
From: cisco-nsp-bounces at puck.nether.net
[mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of Kristofer
Sigurdsson
Sent: Tuesday, September 06, 2005 1:18 PM
To: cisco-nsp at puck.nether.net
Subject: [c-nsp] OSPF scalability

Hello,

We are about to implement a broadband aggregation "stack",
built from a few 7301's.  The idea is that each of the 7301's
will have "outside" connectivity (despite the wording; no NAT 
involved :-)) and announce via OSPF and iBGP the whole IP range
for our broadband customers.  The boxes will then exchange host
routes (connected routes) internally so that even if the "wrong"
box gets a packet for a customer initially, it delivers it to
the right one.

We were planning on using OSPF to exchange the connected host routes
between the 7301 boxes.  I was, however, wondering about it's 
scalability.  Can I realistically expect OSPF to be happy with several
thousand routes, each mapped to a seperate interface (does every 
interface, including the virtual-access interfaces, have to go into the
link state database in every box?).  If not, I guess my best option 
would be iBGP between the boxes?

We were planning to implement this using different OSPF areas, any ideas
on that would be welcomed.

Thanks in advance,
Kristofer

_______________________________________________
cisco-nsp mailing list  cisco-nsp at puck.nether.net
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

_______________________________________________
cisco-nsp mailing list  cisco-nsp at puck.nether.net
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/