Best practices in using loopbacks for router management in a large network

[admin_2]

Best practices in using loopbacks for router management in a large network (500 devices) and IPs for 400 SC0 interfaces

Hello Folks,

I am currently in the process of designing a large campus switched and routed network. At the present time all active interfaces are derived from a

(class-B) /16 VLSMed into /24s, /23 , /22 etc. I will be using all of the ununsed subnets from the production network to run this new network in

parallel untill all services are migrated. So I do not have the option to use IP from their current scheme for router and switch management etc.

So I am thinking about using a RFC 1918 address (10.x.x.x) space VLSMed into /32 for assigning loopbacks on various core/distribution routers and have it redistributed connected into the OSPF. I would like to know if there are any thoughts/best practices in having the loopbacks reditribute connected Vs advertising them via ospf given that summarization may not be possible.

Apart from this, there are about 400-500 switches that need thier management(sc0)/vlan1 i/f configured with IP in a seperate VLAN terminating via trunk into distribution router and i am debating the same issue of having them reditribute connected Vs advertising them via ospf. Any thoughts/suggestions would be greatly appreciated.

Thanks.

Cheers,Padhu

[Richard Burts]

Several of the networks I have worked with use the practice of assigning loopback addresses out of 1918 space while physical interfaces are out of other space. This usually works well.

I think your main question is whether to redistribute connected or to have OSPF advertise via network statements. Redistribution will treat the route as external while the network statement will treat the route as internal. My suggestion is to advertise via a network statement. I do not see any advantage in redistribution and do see a potential disadvantage: if you configure any area in your network as stub (regular stub, totally stubby, or NSSA) the externals will not be advertised into that area and that would limit connectivity.

One suggestion I would make is to organize your loopback addresses and assign them by area: for example all devices in area 0 have looopbacks in 172.16.0.0, all devices in area 1 have loopbacks in 172.16.1.0, all devices in area 2 have loopbacks in 172.16.2.0 and so forth. If you do this, then on the area border routers you can configure summarization of the addresses. This will reduce clutter in your OSPF data base and in the routing table.

I do not see any reason not to do the switches in the same way. If you want to be able to identify the type of device by its address you could allocate one subnet in the area for routers and another subnet in the area for switches.

[n2-goes]

Hello,

I agree with Richard answer. I just want to add that I do not like using vlan1 for switches management. I prefer creating a totally separate vlan for network management.

Vlan 1 is used by STP, vtp etc

Nadine.