MPLS — Labels, LDP & Layer 3 VPN

MPLS forwards packets based on labels instead of IP lookups and enables scalable VPN solutions and traffic engineering.

Facts

Point	Fact
Technology	Multiprotocol Label Switching
OSI layer	Between L2 and L3 (Layer 2.5)
Forwarding	Label-based, not IP longest-prefix match
Classification	Traffic grouped in FEC (Forwarding Equivalence Class)
Label size	32-bit header (20-bit label + TC + S-bit + TTL)
Label scope	Locally significant
Operations	Ingress push · core swap · egress pop
Penultimate hop popping	Standard – second-to-last router removes the transport label
Control plane	IGP (OSPF/IS-IS) + LDP (TCP 646, discovery UDP)
Data plane	CEF + LFIB
Router types	CE (no MPLS), PE (VRF + MPLS + MP-BGP), P (MPLS + IGP only)
VPN label	Assigned by the egress PE; transport label through the core; typical 2-label stack
Route distribution	MP-BGP (VPNv4)
RD / RT	RD makes overlapping prefixes unique; RT controls import/export

What problem does MPLS solve?

Problem	MPLS solution
Large routing tables	Labels instead of IP lookups
Customer separation	VRF + labels
Inefficient routing	LSPs (Label Switched Paths)
WAN scalability	Central routing in the core

How it solves it

Ingress LER classifies traffic into a FEC
Pushes a label
Core routers swap labels (no IP lookup)
Egress LER pops the label and does normal routing

Roles

Component	Function
CE	Customer Edge – no MPLS
PE	Provider Edge – VRF, labels, MP-BGP (LER)
P	Provider core – label switching only (LSR)

FEC

A group of packets that get the same label and are treated identically – e.g. by destination prefix, QoS class or VPN membership.

MPLS label fields

Field	Description
Label	20 bits
TC	Traffic class (QoS)
S-bit	Bottom of stack
TTL	Loop prevention

PHP

The penultimate router pops the transport label – reduces load on the egress PE, controlled via implicit-null label 3. PHP is the MPLS default.

LIB vs LFIB

Table	Function
LIB	All known labels (label information base)
LFIB	Actually used to forward labeled packets (swap/pop/forward)

LDP

Distributes labels between neighbors – UDP 646 hellos to 224.0.0.2 for discovery, then a TCP 646 session for label exchange; builds LSPs based on the IGP.

MPLS Layer 3 VPN (very important)

Combines VRF (routing separation), MP-BGP (VPNv4 routes) and a label stack:

Label	Function
Outer	Transport through the core (LDP) – P routers only know this one
Inner	Identifies the VPN/VRF (learned via MP-BGP, assigned by egress PE)

RD: format ASN:nn or IP:nn – makes VPNv4 prefixes unique. RT: controls import/export of VPN routes in MP-BGP.

MPLS vs classic IP routing

Feature	IP routing	MPLS
Lookup	Longest prefix match	Label
VPN	ACL / GRE	VRF
Scaling	Limited	High
Core routing	Complex	Simple

Special considerations

The IGP must work first
MPLS runs only in the provider core; CE routers don’t know MPLS
MTU must be higher (labels add overhead)

Configuration

ip cef
mpls ip
!
interface GigabitEthernet0/0
 mpls ip
!
mpls ldp router-id Loopback0 force

Verification (CCNP)

show mpls interfaces
show mpls ldp neighbor
show mpls forwarding-table
show ip cef
show bgp vpnv4 all
show vrf

Extras: MPLS TE (RSVP-TE), Fast Reroute (FRR), QoS awareness, carrier-grade VPNs, foundation for SD-WAN.

Tagged ccnp, mpls, wan