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. And see here for current known issues.

New Member

Question on theory of IP assignment...

Question regarding size of subnet to use in this paticular model....

Let say I've got an air-gap between a router, and a PIX, and I essentually need just two IPs to connect the two devices; obviously a .252 subnet with 2 hosts would be appropriate. However, in my earlier days, I used to stick an entire class-C non-routeable network in such a place, with 254 hosts.

I did this until I ran into a problem one day. I had a P-to-P WAN link with two IPs in use for each side, assigned out of a class-C. Some mis-configured equipment started to scan the class-C network assigned to the P-to-P link, and drove the bandwidth util through the roof. A packet debug on the router showed that there were tons of ARP requests for hosts that didn't exist; remember only two out of the 254 were in use.

Question: To avoid having this happen, is it best to keep the size of your subnet that you assign, to a minimum, to keep the broadcast domain low? In other words, are there reasons other than address conservation, to assign small subnets where you only need a few IPs. Thoughts? Comments?



Re: Question on theory of IP assignment...

You should only use the /30 networks with your serial links. The fact that a misconfigured device tried to scan every IP address on the network was the problem. What if the device tried to scan for a network that was a class A subnet /8 Would that still be your fault for having that device trying to find every IP address in that range :-)

New Member

Re: Question on theory of IP assignment...

One other benefit to keeping the address range small is that it is more efficient and easier to perform route summarization(or aggregation if you like) in the long run. Obviously, for aggregation, we need to use contiguous addresses. (I will throw out two arbitrary examples involving 4 P to P links for illustration).

If you take the scenario of a large address space say and assign networks for these 4 links, say:,,,, etc. This leaves a large unused space. And if you try to aggregate, even though only a few addresses are used, the routing table will contain 4 routes since we can not aggegate these four subnets.

If you take the scenario of starting with a small addess space for the four links, we could use This would yield subnets:,,, and These routes could then be aggregated into three routes:,, and Granted three routes versus four routes is not a big difference. Yet if we are trying to aggregate for potentially dozens or more routes for bigger sites, aggregation can benefit routing table size very quickly.

CreatePlease login to create content