The MPLS working
group is addressing the issues of the scalability of routing, the
provision of more flexible routing services, increased performance,
and more simplified integration of layer 3 routing and
circuit-switching technologies, with the overall goal of providing a
standard label-swapping architecture.
MPLS introduces a
new forwarding concept for IP networks. The idea is similar to that
in asynchronous transfer mode (ATM) and frame relay networks. A path
is first established using a signaling protocol; then a label in the
packet header, rather than the IP destination address, is used for
making forwarding decisions in the network. In this way, MPLS
introduces the notion of connection-oriented forwarding in an IP
network. MPLS thus offers a new solution for directing the traffic
along the computed paths-a significant requirement for traffic
engineering, establishing a path and sending traffic along that
path. This provides the network engineer with a level of
functionality equivalent to what virtual circuits provide in ATM
networks. In the absence of MPLS, providing even the simplest
traffic engineering functions (e.g., explicit routing) in an IP
network is very cumbersome.
The following is
a very brief introduction to MPLS. Two signaling protocols may be
used for path setup in MPLS:
Label Distribution Protocol (LDP) and
extensions to RSVP.
The path set up
by the signaling protocol is called a label switched path (LSP).
Routers that support MPLS are called label switched routers (LSRs).
An LSP typically originates at an edge LSR, traverses one or more
core LSRs and then terminates at another edge LSR. The ingress edge
LSR maps the incoming traffic onto LSPs using the notion of a
forwarding equivalence class (FEC). An FEC is described by a set of
attributes such as the destination IP address prefix. All packets
that match a given FEC will be sent on the LSP corresponding to that
FEC. This is done by prepending the appropriate label to the IP
packet. The core LSRs forward labeled packets using only information
contained in the label; the rest of the IP header is not consulted.
When an LSR receives a packet it looks up the entry in its label
information base (LIB), and determines the output interface and new
outgoing label for the packet. Finally, the egress edge LSR will
remove the label from the packet and forward it as a regular IP
packet. Naturally, this description omits many of the subtle
details, but they are beyond the scope of this section. The MPLS
signaling protocols used for traffic engineering are described in
Figure 9 - A
simplified LSR forwarding engine