Segment routing

Operators can incrementally add the SR features they need to an existing LDP/RSVP network with the options they’re comfortable with. This eases the migration to SR and enables them to gain operational experience before introducing more powerful features.

The graphic above shows an incremental approach to deploying the principal applications of segment routing.

Shortest path routing is a good starting point for introducing SR as a replacement for LDP to enable better protection coverage. It only requires control plane extensions for IS-IS or OSPF. Global SR label blocks and node and link adjacencies can be configured for each router, after which SR-IS-IS or SR-OSPF tunnels can be used by IP services. 

Constraint-based forwarding could be a logical next step because it enables more granular programming of the network, for example, to engineer low-latency paths for delay-sensitive applications, or to ensure that traffic flows are kept within a controlled set of links (data sovereignty). 

By augmenting the SR shortest path with an IGP flexible algorithm (Flex-Algo), it becomes possible to add topology constraints and alternative link metrics to the shortest path calculation. This application invokes SR traffic engineering capabilities (SR Policy and Flex-Algo) and programmatic control, for example, with BGP and BGP-Link State (BGP-LS).

SR-TE capabilities can be further extended with BGP and BGP-LS to include BGP prefix and peer segments, and steer traffic to a particular egress point (i.e., egress peer engineering).

Path diversity and end-to-end protection enable highly available premium services as a more scalable alternative for MPLS fast reroute based on RSVP-TE. SR-TE LSPs with diverse primary and secondary paths are enabled by including Shared Risk Link Group (SRLG) constraints for the candidate paths.

Seamless Bidirectional Forwarding Detection (BFD) is used to quickly detect failures, including silent failures that are not visible to the control plane, and to trigger a failover to the secondary path if the primary path fails. 

Peering engineering enables a headend router to steer traffic across a specific downstream peering link between two IGP domains. This is useful for applications such as traffic optimization, resiliency or load balancing across links that interconnect domains.

SR is an ideal technology for engineering forwarding paths with granular policy constraints. These constraints can include the nodes and links to include in, or exclude from, the path, physical diversity, administrative state and path metrics such as available bandwidth, accumulative latency and maximum number of hops. 

SR takes a source-based routing approach that only requires the ingress or headend router to maintain policy and state information about the path. In its simplest form, a segment route is a sequence of segments that must be traversed when forwarding packets along a constrained path to meet a given policy.

Nokia began supporting SR on routing platforms in 2015. We now offer a robust, comprehensive and versatile SR toolkit that has been validated and deployed by network operators for a variety of applications. 

Protection, restoration and assurance of SR connectivity are supported by a range of mechanisms. Loop-free alternate (LFA ) provides a pre-computed path to be used in case of a local failure. Refinements to the basic LFA capability, such as topology independent LFA (TI-LFA), remote LFA (RLFA) and directed LFA (DLFA) provide strong coverage for various topologies and minimally impact service restoration.  Seamless Bidirectional Forwarding Detection (SBFD) complements LFA and enables rapid and deterministic end-to-end path failure detection in as little as 30 ms.

Programmatic control of traffic-engineered path installation in headend routers can be achieved in various ways—for example, using the Path Computation Element Protocol (PCEP) from an overarching centralized Path Computation Element (PCE). Other distributed mechanisms include BGP, BGP-Link State (BGP-LS) and model-driven CLI (NETCONF/YANG).

Traffic engineering of paths can be achieved using SR-TE, SR-Policy or Flex-Algo. From a user perspective, SR-TE is similar to traditional traffic engineering mechanisms such as RSVP-TE but offers much greater scalability and a natural migration path. SR-TE paths can be computed locally or controlled by a PCE. SR Policies consist of multiple candidate paths. One of these paths is activated at a given time, and equal cost multipath (ECMP) is used to load-balance traffic.

If distributed path computation meets a service provider’s operational objectives, IGP Flex-Algos complement SR-TE by distributing prefix segments with specific optimization objectives and a wide range of standard and custom constraint metrics that can be advertised by supporting router nodes.

The control plane for SR is responsible for the advertisement of SIDs. The control plane for SR-MPLS can be IPv4 or IPv6, while SRv6 uses an IPv6 control plane. SIDs can be advertised statically, or more commonly by using IS-IS, OSPF or BGP protocols with extensions for SR and TE.

If a PCE is used, it will listen to the IGP to discover the topology. A prefix SID identifies the node on which a route can be found and is globally unique in the network. Adjacency SIDs are assigned to individual links on the nodes and are locally significant. A prefix segment is typically an ECMP-aware multi-hop path while an adjacency segment, in most cases, is a one-hop path. Other SID types include Anycast SID, a prefix advertised by a group of nodes known as an anycast set, and Binding SID, which can be used to steer traffic into a policy and support inter-domain transport tunneling.

The data plane can be based on MPLS (for SR-MPLS) or IPv6 (for SRv6). MPLS over User Datagram Protocol (UDP) can also be used for paths that transit IP-only network equipment. When deployed over an MPLS data plane, SIDs are allocated from a reserved block in the MPLS label space. To advertise segment reachability information within IPv4 or IPv6 routing domains, SR-MPLS and SRv6 can use IGP extensions for OSPF and IS-IS. This enables a straightforward migration that preserves all transport capabilities of LDP and RSVP-TE while improving failure recovery.