Frame Relay (legacy WAN): DLCIs, LMI, point-to-point and multipoint
In short: Frame Relay was the main shared WAN service of the 1990s and 2000s. Sites joined a provider network with one serial port and reached other sites over permanent virtual circuits identified by locally significant DLCIs. It is now legacy, replaced by MPLS, Metro Ethernet, internet VPNs and SD-WAN, but its ideas still appear in exams and in how routing protocols handle non-broadcast networks.
Why learn a legacy technology?
Frame Relay is legacy. Providers have switched their networks off almost everywhere, and newer Cisco software no longer focuses on it. It was replaced first by MPLS VPNs and Metro Ethernet, and today by internet links joined with IPsec VPNs and SD-WAN. It is still worth a short look because older exam material, old router configurations and textbooks use it, and because its NBMA (non-broadcast multi-access) behaviour explains several routing protocol features, such as OSPF network types, that you still meet in DMVPN and other overlay designs.
How Frame Relay worked
Each site had one serial link (the access circuit) to the provider's nearest Frame Relay switch. Across the provider network, virtual circuits joined pairs of sites. Most were PVCs (permanent virtual circuits) set up by the provider in advance.
- DLCI (data link connection identifier): the number that identifies a virtual circuit on your local access link. It is locally significant: head office might use DLCI 102 to reach a branch that uses DLCI 201 for the same circuit.
- LMI (Local Management Interface): keepalives between your router and the provider switch, which also report whether each PVC is active. Types were Cisco, ANSI and Q.933A; Cisco routers auto-detected them.
- Inverse ARP: lets a router learn the far-end IP address on each DLCI automatically, building the IP-to-DLCI map.
- CIR (committed information rate): the speed the provider promised per PVC. When the network was congested it set FECN/BECN bits to warn senders, and frames marked DE (discard eligible) were dropped first.
Building a lab Frame Relay switch
In labs, a spare router acted as the provider's switch. It switched frames between interfaces based on the incoming DLCI, acted as the DCE and supplied clocking and LMI. The core idea, on older IOS releases, looked like this:
frame-relay switching
interface serial 0/0
encapsulation frame-relay
frame-relay intf-type dce
clock rate 64000
frame-relay route 102 interface serial 0/1 201
interface serial 0/1
encapsulation frame-relay
frame-relay intf-type dce
clock rate 64000
frame-relay route 201 interface serial 0/0 102
Point-to-point and multipoint designs
A router could use its physical interface or sub-interfaces. The choice mattered because of split horizon: a distance-vector protocol will not advertise a route back out of the interface it learnt it on. On a hub with one multipoint interface, routes from Branch A could not be passed on to Branch B.
| Design | How it works | Effect |
|---|---|---|
| Point-to-point sub-interfaces | One sub-interface and one subnet per PVC | Behaves like separate leased lines; no split horizon issue |
| Multipoint (physical or sub-interface) | One subnet for many PVCs, IP-to-DLCI map per neighbour | Saves addresses but needs split horizon disabled or OSPF network types chosen carefully |
interface serial 0/0
encapsulation frame-relay
interface serial 0/0.102 point-to-point
ip address 10.0.12.1 255.255.255.252
frame-relay interface-dlci 102
interface serial 0/0.200 multipoint
ip address 10.0.200.1 255.255.255.0
frame-relay map ip 10.0.200.3 103 broadcast
The broadcast keyword let routing protocol multicasts be copied to each PVC. On multipoint networks, OSPF needed a suitable network type (non-broadcast with neighbours listed, or point-to-multipoint), and the hub had to be the designated router.
Real-world example: retiring the last circuits
A regional builders' merchant ran twelve yards on a Frame Relay hub-and-spoke network until its provider announced the service would close. The network team replaced it in stages with an MPLS VPN, mapping each PVC to an MPLS site, and later added broadband with SD-WAN. The lesson they recorded was that the old OSPF configuration contained Frame Relay-specific settings (neighbour statements and network types) that had to be removed during migration, or adjacencies failed on the new Ethernet links.
Practice
What does it mean that a DLCI is locally significant?
Why did multipoint hubs cause problems for distance-vector routing protocols?
Name two technologies that replaced Frame Relay.
Key takeaways
- Frame Relay is a legacy WAN; MPLS, Metro Ethernet and SD-WAN replaced it.
- PVCs were identified by locally significant DLCIs; LMI reported their status.
- Inverse ARP mapped IP addresses to DLCIs automatically.
- Point-to-point sub-interfaces avoided split horizon problems on hub sites.
- Its NBMA ideas still explain OSPF network types used over overlays.
Original content written by Mohamed Fayas for FM TecHub. © 2026 FM TecHub. You're welcome to share a link to this page; please don't copy the text elsewhere without permission.