1. Home
  2. Cisco Systems
  3. OSPF newbie - multi-site design and Qs

OSPF newbie - multi-site design and Qs

I am in the process of learning-for-design-and-deployment OSPF. Here is the existing setup (based on a recent migration from Nortel to Cisco):

Corporate headquarter with two 6509s in HSRP mode, which play also the rols of inter-VLAN router ("s", if you think at one as passive, though, of course). The two have uplinks to a 3845, which connects (terminates) point-to-point via full and chanelized (sharing some channels with voice) T1s seven other locations. At the other ends I have 2825s, and, behind those, either 650xs or 450xs, doing inter-VLAN routing for those remotes.

Right now everything is statically routed (we opted for static routes in the first phase of migration, as we prefered the comfort of knowledge of such, during vendor change, in regards to total unfamiliarity with Cisco stuff, in the beginning). We are in a hub-and-spoke environment, thus al remotes connecting through the core router (3845) no direct connections between any remotes. Most of the remotes have dual or tri T1s connected back into the 3845 (with 7 sites we have filled the 3845 with plain multi-T1 as well as drop-and-insert (for data+voice) cards!)

I think it is time to move to a routing protocol, and OSPF seems to be a good choice, considering that it is also supported on some non-Cisco firewalls we have, which I plan to expand this to, in a later phase (site-to-site VPNs).

With the above background (sorry for the length - shooting for clarity), here is my question: what would be the pros and cons of the following options (if I correctly understand OSPF):

  1. area 0 on the ethernet interfaces of the 3845 connecting to the 6509s at the HQ, and also including those 6509s, then separate area for each WAN link + internal remote LANs

  1. same as above, but the external (serial) links of the 3845 also in area

0 (and the rest separate areas)

  1. all sites in one area 0

Comments or pointers to similar designs would be greatly appreciated. Any other options?!?

TIA, Papi

Reply to
papi
Loading thread data ...

It depens how many boxes you have. But I think its good to keep this "area

0" in LAN, which is stable. Not in WAN, so there is much less re-calculating.
Reply to
Make

you need to limit which VLANs have OSPf adjacencies on them - you only need a couple. These VLANs are then candidates for carrying traffic between the core routers.

The two have uplinks to a 3845, which connects (terminates)

this stresses the 3845 as it runs a copy of OSPF algorithm for each attached area.

however, all areas are pretty small, so not a major limitation.

things that limit the size of an area are documented in the RFCs - so read those. and / or find a copy of the OSPF bible - OSPF - Anatomy of a routing protocol by John Moy

things to worry about for area size are: no of routes (1000's, so not an issue for you) no of routers (100+, unless some of your boxes are CPU / memory limited). amount of "loops" inside an area, since LSDB info replicates within the loop during a state change (10s ideally). required convergence time - more info in an area to propagate limits the recovery time after a fault. differences in structure - i.e. backbone should be resilient, other areas may not be. differences in reliability - unreliable links and networks should not be in the same area as reliable bits. area 0 links - if a border router loses all its backbone links, wierd things can happen to traffic thru that box which cause routing problems between areas. area 0 should stay internally connected during a fault wherever feasible.

better for central 3845 - but

my preferred design for this size network, since ABRs massively complicate OSPF.

a lot depends on whether the remote sites can still work without the HQ - the inference from the design (with single central WAN routers) is that they can, or that you are willing to live with single points of failure.

since the WAN is single path, you could make each site a separate OSPF network, and use statics between them.

more manual complications, but more stability if you get flapping links or other soft faults.

Reply to
stephen

Thank you very much for your thorough answer! One thing I mentioned was the fact that each site has at least two T1s going between it and the HQ, right now with the traffic statically routed-distributed between the two. I was wondering how OSPF (perhaps with some QoS enabled) could take care of load distribution and - especially - redundancy (i.e. under which of the three scenarios in regards to division in areas would the redunndancy and even prioritization play better, and why)?

Thx again, Papi

Reply to
papi

Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.