I have a quick question for you.
We are looking to expand our network but with regard to our ever evolving
OSPF environment, an issue has been put on the table and I would like your opinion on this.
We currently have 56 routers in area zero. This number will increase when
we expands your facilities. Some of us feel (myself included) that we should limit the number of routers in area zero. I am concerned that the area zero database will get too large and
smaller routers will become increasingly sensitive (with their limited CPU capacity and the large size of the area zero database) to changes in the area zero database due to link state changes originating from the other areas.
Some feel that the impact to area zero can be lessened by separating
area zero from one subnet per location to multiple subnets per location.
They feel that dividing area zero into multiple subnets will reduce the
database size and the associated size related issues that are more
prevalent on smaller routers.
I feel that the area zero database stays the same size regardless of how
many subnets you divide the area into.
I would like your opinion on this subject. Specifically, how
due we determine the size of area zero (number of routers)? Will
creating multiple area zero subnets help? Should I be concerned with the
size of the area zero database?
I think you are correct with regard to the data base size of area zero.
Personally, I would look to reduce the number of routers in the backbone. Is thre any way you could migrate some of the routers to different areas ?
Adding extra subnets should actually make the ospf database 'larger' : as numbers of Link State descriptions are increrased = a larger OSPF database.
The following URL should help a bit when designing your OSPF areas :
I have a similler situation but have 1601 routers in the area and old 2511,2501 routers
some times it is not working good and other times it works better than enough
could u tell me the reason for that???
Have you considered route summarization? This may help to reduce your table size and reduce the CPU usage by minimizing the amount of times the SPF algorithm is calculated.
Can you do a route summarization inside the area 0 backbone. I thought that you can do it only with ABRs, which are connected to two areas.
But for the above question I would suggest to split an area 0 into smaller areas. It will be less routes inside the backbone.
Here's my opinion, for what it's worth...
The size of Area0 is heavily dependent on the amount of RAM you have installed on all the routers in Area0, whether the CPU can keep up with everything (primarily during convergence from a topology change), and how much time for convergence you can handle. The more routers there are in a given area, the busier your routers' CPUs are. The more often a topology change occurs (causing reconvergence), the more the CPUs crunch their own network picture. The busier the router is, and the more memory it is using for its various tasks, the longer it takes to reconverge.
I've heard the rule of thumb that 50 routers is the max you want in an area. If your at 56 routers now, I beleive you *really need* to split now, before you even begin expanding your network.
Another thing to consider is placement of your DR and BDR(s). Make darned sure your DR is the beefiest router you've got, in the best connected, most reliable part of the network. BDRs should be in strategic locations to be able to pick up DR duties when a failure severs it from the DR.
OSPF uses a variety of tables and databases. The number of routers and their relationships determine the size of the neighbors database, but you still need to worry about the routing table itself, which will grow as new networks are created, and as redundant/duplicate paths to the networks are created.
Creating multiple subnets will only aggrevate the situation. Also, don't confuse OSPF areas with IP networks/subnets. Route summarization is a strength of OSPF, so even within a single area, a single router summarizes its routes in order to advertise them.
Regarding the size of the Area0 database(s), you need to consider all of this above. The design of your OSPF network is absolutely critical in making sure you can recover from network failures quickly. An improperly laid out scheme can cause slow convergence, or even worse, *perpetual* or constant reconvergence.
Not sure if you need the help, but I've been through this a couple times. If you'd like to collaborate a bit more, email me at firstname.lastname@example.org
Hope that helps!