Summary

Traffic engineering is a set of techniques to steer certain types of traffic via a path different from the IGP-calculated shortest path.

Since link-state IGPs provide a view of the entire topology (IS-IS level or OSPF area), routers can calculate paths with various constraints. One useful example of this was TI-LFA where the backup path was calculated with the constraint of excluding the protected link/node/SRLG. For traffic engineering, it is also possible to request custom paths excluding certain links, or using other constraints like available bandwidth or delay.

In Segment Routing, routers feed network topology information to a controller using BGP-LS, the controller computes routing policies and installs them into relevant routers using BGP-SRTE. The advantages of using BGP to distribute network topology information and routing policies is that it’s a standard protocol supported by all routers, and the network designer can leverage an existing BGP topology to propagate traffic engineering information to all routers, instead of having the controller connect to each router separately.

Automated steering

Each SR-TE policy has a color, which can be used to steer traffic into that policy. A BGP route with color extended community matching the SR-TE policy color will be mapped to that policy.

Consider the topology:

There is an MPLS L3 VPN service between R1 and R8. Design requirements are that traffic for 10.0.0.0/24 is sent strictly over the yellow links. Traffic Dictator calculates an SR-TE policy with color 101, and L3 VPN routes matching endpoint 8.8.8.8 and color 101 will be mapped to that policy.

Emulating Automated Steering with legacy and budget routers

One of the goals of Traffic Dictator is to lower the entry barrier for Traffic Engineering, so that network operators are not locked in with expensive routers from a handful of big vendors, but have a wider choice of various routers available to them, including whitebox routers and open source routing implementations.

Some of these devices don’t support BGP SR-TE or PCEP, but almost any router that supports basic MPLS, also supports BGP Labeled Unicast. This allows network operator to use BGP-LU to install Traffic Engineering policies. BGP-LU doesn’t have a concept of “color” like SR-TE, but Traffic Dictator offers an option of “service-loopback” to emulate the SR-TE color functionality with BGP-LU.

In this topology, R8 has a “service-loopback” 172.16.1.1. It is not advertised into IGP but advertised into BGP-LU with nexthop of 8.8.8.8 which is the main R8 loopback reachable via IGP. R8 advertises 172.16.1.1 to R1 with a lower local preference, as a backup route to preserve best effort path if the controller fails. When advertising the L3 VPN route for 10.0.0.0/24, R8 sets nexthop to 172.16.1.1.

Traffic Dictator sends a BGP-LU route to R1, with prefix 172.16.1.1, and nexthop of R3 interface address. This way, the same behaviour as automated steering can be achieved even if routers do not support BGP SR-TE.

Note that label stack in this case will be just <R8> and not <R3, R8> – this is because unlike SR-TE, where the first label is used by router for nexthop resolution, BGP-LU route already has an IP nexthop so there is no need to send a label of a directly connected router.

Further information

Check out Traffic Dictator White Paper on the Products page or email info@vegvisir.ie with any questions.