OSPF router-id and IP addressing best practice ?

[ivanbarkic]

Hi,

I'm doing large enterprise network which includes new IP addressing
and OSPF as routing protocol. I took 10.0.0.0/8 for clients, servers
and other stuff in network and divided it hierarchically so that I can
do summarizing later on. My WAN connections are all metro ethernet L2.
For WAN IP addresses I took 192.168.0.0, simply for easier distinguish
between WAN IPs and other IPs. WAN addresses I will be able to
summarize. My dilemma is router loopback's IP addresses. I'm planning
to use 192.168.255.0, 192.168.254.0, 192.168.253.0, etc, for all
loopbacks in my network and those subnet will not be included in
summarization. Reason for that is also easier distinguishing so I know
that those subnets are my loopbacks.
At the end, one day, I could have almost 1200 routers in all OSPF
areas. Is it good idea to put loopbacks like that? I will have 1000
lines in routing table without summarization, just from loopbacks .
If I use loopback from C classes which can be summarized, I won't be
able to easly recognize loopbacks in my routing table. That loopbacks
will be management addresses and used for accessing and for mgmt
tools.
What is best practice for loopbacks globally in enterprise network?

Another question is about OSPF router id. Is it smart to assign router
ids like, lets say, 1.1.1.1, 5.5.5.5, since those addresses won't
exist in my network? Or, is it better to use Loopback IP address?

Thanks.

[Richard Burts]

Ivan

The logic of choosing 10.0.0.0 for subnets for end stations and choosing 192.168.0.0 for WAN is clear and makes good sense. I do not understand your logic in choosing 192.168.255.0 for loopback interfaces.

I would suggest that it might make much better sense to choose something like 172.16.0.0/16 for loopback addresses. If it is safe to assume that you would not have more than 255 OSPF areas and that no area would have more than 255 OSPF routers, then you might use a method in which the third octet is the OSPF area number and the fourth octet is for individual routers within the area.

Unless there is some reason that is not obvious in your post I would recommend using the router loopback address as the OSPF router ID. That is simple and consistent and will scale well. I will sometimes use OSPF router IDs like 1.1.1.1 or 5.5.5.5 when I set up something in the lab so that they are very unique and recognizable. But that does not scale if you are going to have 1200 routers.

HTH

Rick

[ivanbarkic]

Well I cannot use 172.16.0.0/12 since it is in use in existing IP scheme. The reason why I choose 192.168.255.0 is simple: to easily recognize loopback among other IP addresses and cause it is last octet.