Problem solve Get help with specific problems with your technologies, process and projects.

Using subinterfaces on a Cisco router

When configuring a router for Frame Relay, do you need a serial sub-interface couple to the primary interface?
No. You can use the primary interface if you want to. It's just not a very good idea. There are a lot of reasons; let's cover a few of them:

Reason - Inverse ARP
It is well known that on Ethernet, you use ARP to determine the IP to MAC address mapping, i.e. you have the destination IP address and you want to send the packet to a PC on Ethernet, so you send an ARP request to find out who has the destination and would they please send you their MAC address. Now you have the MAC address, you can send the Ethernet packet.

Now, the PC that receives the packet knows the sender's MAC address, but not necessarily the sender's IP address. The receiver will have to reply to the packet that came in, to confirm it got the packet OK. The receiver will perform the opposite, known generally as Inverse ARP. The receiver sends the Inverse ARP request and maps the MAC address to an IP address. The receiver can dispatch the confirmation packet.

But for Frame Relay, there is no ARP on physical interfaces. There are quite a few reasons for this, which I won't cover here. Thus the router would not know which PVC to send the packet to. So on a physical interface for Frame Relay you manually configure static maps that map the remote IP address to a PVC. Creating static maps is prone to error, hard to read, and hard to administer.

On a sub-interface, however, you can send ARP and inverse ARP, that is, I have the DLCI, what is your IP address, to determine what the remote IP address is.

Reason - Interface Administration
Consider the following scenario, you have a major Frame Relay hub, with 10 PVCs to the spoke sites. If you use the primary interface only, you will have 10 PVCs attached to the one interface. You then need to create mapping statements for each PVC. This is hard to maintain as time goes by. If a single PVC goes down, it is very hard to detect when using physical interfaces.

It is much better to have a different interface for each PVC, and we use sub-interfaces to do this. It makes sense to most people. You can see quickly if a PVC is down because the interface is down. Once the interface is down, OSPF, EIGRP and ISIS routing protocols can converge on a backup path much quicker, if there is one.

Reason - Performance Management
Many performance management tools, like MRTG ( www.mrtg.org), work best when each PVC has an interface and an IP address. Remember that a physical interface can't have a separate IP address for each PVC. Only one IP per interface is allowed.

Reason - Broadcast Control
Keep in mind that the physical interface can have only a single address, so your static map statements have to be in the same network. When you need to broadcast the broadcast goes across every PVC !! This is usually very bad.

Reason - Routing Protocols - Split Horizon and Neighbor Discovery
This is the biggie. If you have use the physical interface, distance vector routing protocols like RIP have split horizon problems. The Link State routing protocols like OSPF have to use special mechanisms to discover their neighbors, and need special configuration.

You might see physical interfaces used in very specialized situation. If you do, you generally know why they are there. Using physical interfaces was the only way to configure Frame Relay for a long time and can be seen in older networks. Sometimes people just don't know about sub-interfaces. Frame Relay took a long time to catch on in the United States compared to Australia and Europe.

As a rule, always use sub-interfaces.
This was last published in May 2001

Dig Deeper on Network Infrastructure

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.