455 Words. Plan about 3 minute(s) to read this.
We soldier on with MPLS architecture basics, covering now packet forwarding and label switched paths (LSPs), to complete the picture of how labeled packet is forwarded from one part of an MPLS domain to another. So to the bombastic strains of the Dvorak Piano Concerto Opus 33, let us continue.
MPLS Packet Forwarding & Label Switched Paths
- A packet enters the MPLS network at an ingress LSR, logically exiting at an egress LSR. From ingress to egress LSRs, we have what is called the Label Switched Path (LSP). (Which sounds like the Golden Path of Dune mythos. Or not. But I digress.)
- An LSP then, is the set of LSRs a labeled packet travels through to get to the egress LSR of a certain FEC. An LSP is unidirectional, thus implying that traffic flowing in the opposite direction between an ingress/egress LSR pair will use a different LSP. (Not to be confused with LSD, which reportedly makes everything different. But I digress, again.)
- An LSP is considered to be connection-oriented, since the path is built before any traffic actually flows. The connection information is based on topology information, not on traffic flow requirements, thus the path is always there.
- As a packet hops from LSR to LSR along the LSP, each LSR will swap the incoming label for an outgoing label, right up until the packet arrives at the LSR.
- LSRs maintain 2 MPLS-related tables:
- The Label Information Base (LIB) or Tag Information Base (TIB, Cisco-only terminology) has in it all labels assigned by the LSR, and the mappings of these labels to labels received from neighbors. The labels are distributed via label distribution protocols, to be discussed later.
- The Label Forwarding Information Base (LFIB) or Tag Forwarding Information Base (TFIB, Cisco-only terminology) contains the labels actually in use for packet forwarding by MPLS. Different neighbors will send labels for the same IP prefix, but not all of those lables will get used by the LSR, you see.
Other MPLS Applications – peer-to-peer VPN and traffic engineering are applications made possible by MPLS.
- MPLS applications all use the control plane to establish the label switching database (i.e. the Label Forwarding Information Base in the data plane).
- MPLS IP routing control
- MPLS Multicast IP routing control
- MPLS/VPN routing control
- MPLS Traffic Engineering control
- MPLS QoS control
- MPLS applications all use a common set of components, the same as the IP routing applications.
- A FEC definition database for the application.
- Control protocols that exchave the FEC table between LSRs.
- A control process that binds labels to FECs and exchanges these bindings between LSRs.
- An optional internal database of FEC-to-label mappings.
- Each application uses its own protocols to exchange FEC tables and/or FEB-to-label mappings between LSRs.