Cisco Support Community
Showing results for 
Search instead for 
Did you mean: 

Welcome to Cisco Support Community. We would love to have your feedback.

For an introduction to the new site, click here. If you'd prefer to explore, try our test area to get started. And see here for current known issues.

New Member

ospf design issue

In one ospf area can contain how many routers?

In one AS which use ospf as IGP can contain how many routers?

I have seen the document about it on cisco web site before,but I can't find it now.

Can anyone find it??


Re: ospf design issue


Re: ospf design issue

The first URL is way out of date, I'll have to talk to the CCO guys about changing it.... If you are running a network of 800 series routers, you might not get more than 50 routers in the network before your convergence times become unacceptable, but.... :-)

The first thing to ask is, really, what sort of convergence times are acceptable. The less you care about convergence times begin really fast, like under one second, the more you can stretch the bounds of a single area. So, begin by deciding how fast you want the network to converge.

The second issue is that the number of routes in the area is going to have as much impact on convergence as the number of routers--in fact, it will probably have a larger impact than the number of routers. You could have 20 routers in an area, each one originating 5,000 routes, and the area is likely to converge very slowly. The actual SPF run times would be quick, but installing 100,000 routes in the routing table, and then getting the CEF and hardware switching tables updated, can take some time.

If I wanted moderate convergence, on the order of 9 or 10 seconds (the fastest you can get with the default timers, anyway), I'd feel pretty safe about a couple of hundred routers, with 5,000 routes or so total in each area. If I wanted to converge more quickly, and I was willing to tune the timers, then I'd want to reduce my area size somewhat, depending on the convergence speed I wanted. I'd also watch the types of routers in the area--800's are much slower than 7200's, and there's just so fast an 800 can go.

It's also important to remember there are three times when convergence occurs, at initial startup, single events, and traumatic events. I'd try to plan for the one that is most likely to worry me. In other words, can I live with the network taking 20 or 30 seconds to converge if all the routers come up at once (not likely to happen anyway) to get into a larger area size, knowing that single link flaps and such are likely to happen more ferquently, and take much less time to converge? How often could a single link flap take out a very large number of routes (traumatic event)? What sorts of convergence times do I want there?

I'm hoping the new edition of Advanced IP Network Design, currently being written, is going to cover some of these issues, and try to demistify this to some degree. No-one can really give numbers without specifically testing a given topology, with a given set of routers, etc., but you can get a feel for it if you've worked with it long enough.