Benefits of sub-interfaces?

[bubarooni]

What are the benefits of configuring subinterfaces on a serial interface?

I have a router that has two other routers connected to it through Frame Relay. All are currently connected by a fractional T1. The two remote routers are being changed from fractional t1's to 56k dds service. I am getting the new WIC 1DSU 56k4 cards in a couple of days so need to do some reconfiguration on the routers. The Serial interface on the head end router is currently configured as such:

interface Serial0

ip address 10.0.1.1 255.255.255.0

encapsulation frame-relay

no fair-queue

service-module t1 timeslots 1-2

frame-relay map ip 10.0.1.2 17

frame-relay map ip 10.0.1.3 18

frame-relay lmi-type cisco

Seems very short and uncomplicated to me.

However...

I am being advised by the Cisco ConfigMaker utility to change it to:

interface Serial 0

no shutdown

no description

service-module t1 clock source line

service-module ti data-coding normal

service-module ti remote-loopback full

service-module ti framing esf

service-module ti linecode b8zs

service-module ti lbo none

service-module ti remote-alarm-enable

no ip address

encapsulation frame-relay

frame-relay lmi-type cisco

interface Serial 0.1

no shutdown

description connected to a_1750

ip address 10.0.1.1 255.255.255.0

frame-relay interface-dlci 17

interface Serial 0.2

no shutdown

description connected to b_1750

ip address 10.0.2.1 255.255.255.0

frame-relay interface-dlci 18

That is a lot more stuff! What will all this additional stuff do for me? I am a little concerned as I will be forced to change my ip scheme for the serial interfaces under the ConfigMaker utility recommendations but, if there is a benefit would be willing to do so.

Thanks in advance.

[svermill]

One reason that p-t-p subinterfaces are preferred is that split horizon, by default, is disabled on all other types of interfaces. If your head end router learns of networks on remote site A, it can't advertise them back out on that same interface w/ split horizon enforced. So remote site B doesn't learn of remote site A's networks (and vice versa). Of course, in a simple network, you can just statically enter all of your routes. That doesn't scale well and not knowing that new networks need to be statically added could result in some unnecessary troubleshooting later.

P-t-p subinterfaces solve the whole problem. As for the other stuff, I believe that you are seeing some explicit entries that are already on by default. It is kind of like adding subnet zero in routers running current IOS. It isn't necessary anymore, but it reminds you (and others) of the defaults in place.

[bubarooni]

Well, that set a lightbulb off over my head, though it is only dimly lit at this point.

Right now I have static routes defined on all my routers and no routing protocol enabled on any of them.

The configmaker utility actually added the following:

router rip

version 2

network 192.168.1.0

network 10.0.0.0

no auto-summary

Is this in addition to the subinterfaces going to replace all my statics routes I currently have on the individual routers?

Thanks for your earlier reply.

[svermill]

It certainly could. If you decide the subinterfaces are worth the effort, you could then use a routing protocol and endjoy the benefits of split horizon. Of course, there will be some new overhead on your FR network for periodic routing updates, etc. But with only two networks to route, that will be down in the noise level.

[gwoodyard]

Well said! but what about "no ip split horizon?"

That would not help if the were using IPX though.

[Liam3]

As others suggested there are some issues with split-horizon and routing protocols. Again, they can be worked around but they are not fun. That is one of the two major reasons that I use point-to-point subinterfaces except in cases where it's not possible (because of existing designs being multipoint on devices you can't control to readdress). The second reason is minor to some, but point-to-point interfaces give you better knowledge of the network. The interface goes down when the PVC goes down. With SNMP monitoring packages (like the freeware mrtg) you can monitor traffic on a ptp subinterface. Unless I'm mistaken this is not possible with a PVC on a multipoint. The best you can get is the combined traffic of the multipoint. If the multipoint has 50 sites, one could be saturated and you monitoring could show ~2% utilization.

Bill

[r-roman]

I use subinterfaces for an effective backup.

When an interface "see" a switch frame relay, it changes up the Protocol. If exist a problem in a the middle of the frame relay path, for the router the channel is up, even if some Node of the network is down. In that case, the interface is always up and the sentences for backing up the channel, never activate the backup(for example and ISDN port or another WAN port).

For a WAN port using subinterface and using backup sentences, it detect the falling PVC and activate correctly the backup