| FAQ |
  |
| Networking : Tricky Ones : A Network Redesign |
    |
| Question: | I have a little bit of a problem with our network, which is currently growing drastically. We’re running ospf as our interior routing protocol. We have 140 area 0 routers, which are geographically spread around the country. Wide area flapping can be detrimental to the core area 0, therefore, we’re looking at revamping the design of this network. We’re leaning toward multiple AS and BGP. Your two cents on this approach would be appreciated. |
| Answer: | Well, there is definitely a lot to be said for a clean network design! And I can appreciate that WAN links flapping in Area 0 may be a bad thing.
It’s a bit difficult for me to give a good recommendation without a lot more detail about the network (which usually constitutes consulting!). But let’s think about what’s happening for a moment. OSPF Area 0 is your backbone area. So ALL of the various LSA types are moving across this area and updating the database between all of the neighbors. OK, simple enough. If you move to a multi-AS BGP scenario, effectively you’re telling me that you’re going to break the network up into different summarization points. Are you planning on these eBGP neighbors being fully meshed then to prevent flapping? Are they all going to peer with each other and send routes through as well? That is certainly a valid solution based on what little I know about your network at this point. However, if you’re going to redesign and resummarize your network anyway, let me play devil’s advocate on this. Why not simply reduce the number of routers that are area 0, and start doing area summarization in other places? The point being that if you’re aware of problems anyway, and you’re taking steps to clean things up, what’s wrong with fixing the routing protocol that you have? Start analyzing who your neighbors are within Area 0 -- plot out the routes. Specifically, where the flapping WAN links are in question, what can be done for backdoor routes? OSPF is a fairly resilient protocol, but certainly performs better the smaller the database is. So planning your areas and your summarization points will keep things much more efficient in that regard. And planning around your flapping WAN links will help as well. If you’re running in large fully-meshed OSPF pockets, you may also consider tweaking the flooding parameters of OSPF. By default, OSPF will flood LSAs out all participating interfaces (except the direction learned), which may lead to some instability in this fashion, particularly during flapping. Consider the "ospf database-filter all out" command to prevent flooding on a particular interface. Without planning of this nature, you have the potential to run into the same flavor of problem no matter which routing protocol you’re running. Note: You’ll have to adapt my answer to fill in the blanks of what I don’t know about your network! But in general, clean design works well with either BGP or OSPF. For further reading, check out: http://www.cisco.com/en/US/products/sw/iosswrel/ps1831/ |
| last updated 01.03.2003 | |
| 40 User comments (display) | |
user rating:  | |
| [previous FAQ] [next FAQ] | |
| List FAQ in this category | List categories | List all FAQ | List programs |
|