MPLS Traffic Engineering

Introduction to MPLS Traffic Engineering

A typical like OSPF, computes a lowest-cost path according to its metric. A basic SPF calculation does not automatically include TE attributes such as unreserved bandwidth, so several tunnels can select the same path. When the topology changes, SPF can recompute the route, but it still primarily optimizes the IGP metric.

An IGP metric describes the selected path, but it does not by itself describe current TE reservations. In a large service-provider network, several routes can therefore load the same links while parallel links remain lightly used. Traffic engineering adds explicit resource constraints and operator policy to the path computation. A particular LSP can use a different path even when that path is not the lowest-cost result of ordinary SPF.

MPLS Traffic Engineering (MPLS-TE) is a set of mechanisms for computing and signaling LSPs subject to constraints. A router can use TE attributes, compute a path that meets the request, and establish a tunnel on that path. The decision can include topology, link parameters, and resources available for reservation. An LSP can carry an explicit hop list, a bandwidth reservation, or restrictions based on administrative groups. The ingress router performs the computation for the requested tunnel.

The Three Components of MPLS-TE

MPLS-TE is not a single protocol. A typical solution combines TE attribute distribution, constrained path computation, and LSP signaling.

Information Distribution:
Routers need link attributes beyond the ordinary IGP metric. OSPF-TE and IS-IS-TE extensions can advertise values such as maximum, reservable, and currently unreserved bandwidth. TE routers collect these values in a database of network constraints. OSPF carries TE data in Opaque LSAs, while IS-IS carries it in the appropriate TLVs. The advertised values apply to a particular transmission direction.
Path Calculation:
Using the attribute database, a router selects a path that meets the tunnel requirements. The Constrained Shortest Path First (CSPF) method can remove links that fail the constraints and compute a path through the remaining topology. An example request can require at least 1 Gbps of unreserved bandwidth on every link. After filtering, the selection criterion can still be a TE cost or another metric configured by the operator.
Signaling:
After the path is calculated, state must be established on the successive nodes. This is commonly handled by RSVP-TE (Resource Reservation Protocol, Traffic Engineering). The PATH message follows the computed path and creates path state, while the RESV message checks and establishes the requested resource reservation. The result is a Label Switched Path (LSP). RSVP-TE state is soft state and is refreshed by messages, so a missing refresh can remove the reservation.

1. TE Attributes in IGP Extensions

Ordinary OSPF SPF computation uses the topology and link costs carried in the link-state database. Those values alone do not describe TE reservations or the bandwidth available to a new tunnel. OSPF-TE and IS-IS-TE add attributes intended for traffic-engineering calculations.

OSPF-TE and IS-IS-TE add optional Type-Length-Value (TLV) fields to link-state information. The extensions can advertise, among other values:

Maximum Link Bandwidth: The total configured bandwidth of the link.
Reservable Link Bandwidth: How much of the total bandwidth is available for TE tunnel reservations.
Unreserved Bandwidth: How much reservable bandwidth is currently free.
Administrative Group (Link Color): An administrator can assign "colors" (represented by a 32-bit bitmask) to links to group them logically, for example, tagging all fiber links as "blue" and all satellite links as "red".

The attributes are flooded through the domain as link-state information. A TE-enabled router collects them in a separate Traffic Engineering Database (TED). TED describes the topology known to the router and the attribute values from the latest advertisements, so it changes as the network state is updated.

2. CSPF Path Computation

With TED data, the ingress router at the start of the requested tunnel can compute a path. It uses Constrained Shortest Path First (CSPF), a constrained path-computation method that can combine topology filtering with an SPF calculation.

Before computing the path, CSPF prunes the topology stored in TED. It temporarily excludes links that fail explicit tunnel constraints, such as a bandwidth requirement or an administrative-group rule.

CSPF in Action: A Step-by-Step Example

An administrator wants to create a tunnel from Router A to Router F with a requested bandwidth of $150 \text{ Mbps}$ .

Router A will perform the following CSPF process:

Step 1 (Receive Constraint): The router receives the request for a tunnel to F requiring $150 \text{ Mbps}$ .
Step 2 (Prune the TED): The router consults its TED. It looks at the A→B link and sees that the unreserved bandwidth is only $100 \text{ Mbps}$ . Since $100 < 150$ , this link does not meet the constraint. CSPF temporarily removes the A→B link from the map for this calculation.
Step 3 (Run SPF on the Pruned Topology):CSPF now runs the standard SPF algorithm on the remaining, modified topology. On this pruned map, the only available path is A→C→D→F, which has sufficient bandwidth on all segments.
Step 4 (Return Path):The algorithm concludes that the best constrained path is A → C → D → F and returns this explicit list of hops.

The result is a path that satisfies the bandwidth constraint in the example and the other requirements defined for the service. The 150 Mbps and 100 Mbps values belong to this illustration; an actual decision depends on current TED advertisements.

3. RSVP-TE Tunnel Signaling

Computing the path is the first step. The path must then be established, and the requested resources must be checked and reserved on its successive links. This is the role of the signaling protocol, RSVP-TE.

RSVP (Resource Reservation Protocol) signals receiver requirements and reservation state. RSVP-TE adds mechanisms for traffic-engineered tunnels, including explicit routes and labels. Its exchange uses two key messages sent along the path calculated by CSPF. PATH carries the request in the downstream direction, while RESV confirms resources and labels in the reverse direction.

The PATH and RESV Message Exchange

The PATH Message (Forward Pass):
The ingress router (A) creates an RSVP PATH message. This message contains the explicit path calculated by CSPF (A → C → D → F) and the required resources (e.g., $150 \text{ Mbps}$ bandwidth). It sends this message to the first hop, Router C.
Router C receives the PATH message. It checks whether it can support the $150 \text{ Mbps}$ request on its link to D, records path state, and forwards the PATH message to Router D. Router D does the same, forwarding the message to the egress router, F.
The RESV Message (Backward Pass):
The egress router (F) receives the PATH message and, if the path and policy admit the requested resources, sends an RSVP RESV (Reservation) message upstream along the reverse path. The RESV message can also carry the label that the upstream router should use.
The RESV message arrives at Router D. This confirms the reservation, and D now knows that it should use the label provided by F for packets on this tunnel. Router D then allocates its own label for the tunnel and includes it in the RESV message it sends back to Router C. This process continues backward until the RESV message reaches the ingress router, A.
Tunnel Established:
When ingress router A receives the RESV message from C, the reservation has been accepted on the successive links of that path. The LSP has signaled state, and routers know which labels to use to forward traffic through the tunnel. Router A can now start sending traffic into the MPLS-TE tunnel.