I’ve been really enjoying working with frame relay, but one thing I couldn’t keep straight was point to point and point to multi-point sub interfaces, and the role of the DLCI. For some reason this was a confusing mass that I just could not untangle, so here is what I finally managed to come up with to keep it straight
Point to Point Frame Relay Connections
The problem with it of course, is that it does not scale, requiring one serial interface for each router connection. A full mesh on a reasonable sized network could incredibly complex and even a large hub and spoke will require a great deal of hardware. Each link in the network has to have a serial interface and each serial link has to be on it’s own subnet. (See the /30 IP addresses)
The other thing to remember is that frame relay is a layer two protocol, so IP addressing is not used in the frame relay switch. For the frame relay to work we must remember to use Data Link Connection Identifiers (DLCIs). For each logical connection or circuit, there is one DLCI specified. Because there is only one DLCI, there are no frame relay map statements and no inverse arp required.
Point to Point sub interface, Frame Relay Connection
The point to point sub interface configuration is the same as the configuration above, but we logically segment our serial interface, reducing the hardware requirements. However, like the example above there is some subnetting to be done.
This also uses frame-relay interface-dlci command for each sub interface connection.
Remember that the DLCI is supplied by the service provider, and are locally significant only. They are simply flags that indicate the ends of our virtual circuits.
Point to Multi-Point Frame Relay Connections
With point to multipoint, we have a hub router using a single serial connection to the frame relay switch, but; we have multiple DLCIs specified through either frame map statements for static configuration, or inverse arp for dynamic configuration.
Subnetting can be done or not, depending on the network configuration. In the example here, all three connections are on one subnet.
Frame-relay map <IP address> <DLCI #>
So how do we know what method of connection to use?
Personally, I prefer the point to multipoint as there is less fuss with and fewer IP addresses used. For a small configuration, (maybe up to three serial links) I could see using sub interfaces, but for larger configurations, I think the p2mp wins just on the fewer IP addresses required.
From the Cisco Library:
- Subinterfaces act as default NBMA network
- Can save subnets because uses single subnet
- Good for full-mesh topology
- Subinterfaces act as leased line
- Each point-to-point connection requires its own subnet
- Good for star or partial-mesh topologies
- In Point to Point More IP Addresses are consumed than the Multipoint Subinterfaces.