Segment routing

Segment Routing (SR) uses the principle of source routing. A source node directs a packet through an ordered list of instructions, called "segments". A segment can represent any instruction, such as a topology instruction (follow a path) or a service instruction (apply a service to this packet). Segment Routing thus enables a flow to follow a specific path, without having to create and maintain states on the equipment located along that path. Only the source router maintains a specific state, in the form of a list of segments to be added to the packet. This list of segments is called Segment Routing policy (SR policy). A segment is identified by a segment identifier (SID: Segment ID).

Segment Routing is applied to the MPLS transfer plan (SR-MPLS) without modifying the MPLS architecture or the MPLS transfer plan. A segment is identified by a label. A list of segments is encoded as a stack of labels. The segment currently being processed corresponds to the label at the top of the stack. When the segment is finished, its label is removed from the top of the stack.

Segment Routing applies to the IPv6 transfer plan (SRv6) by defining a new type of routing header. A segment is identified by an IPv6 address. An ordered list of segments is encoded by an ordered list of IPv6 addresses in the IPv6 routing header. The segment currently being processed is indicated in the destination address field of the IPv6 header. The next segment is indicated by a pointer in the routing header.

Segment Routing supports a distributed or centralized control plan.

Generic segments are typically announced by distributed IP routing protocols such as internal link-state routing protocols (OSPF and IS-IS) or inter-domain routing protocols (BGP).

In distributed mode, the Segment Routing policy is calculated by the node wishing to use it. This calculation uses topological and segment information distributed by the routing protocol.

In centralized mode, a network controller calculates the Segment Routing policy and decides which node should use it. A PCE (Path Computation Element) server can fill this role, and in this case uses the PCEP protocol to exchange this information with the source nodes. But other protocols are also possible, such as BGP or NETCONF.

Segment Routing has many applications, which can be grouped into three categories: improved network response to faults (Fast ReRoute, IGP convergence without micro-loops); traffic engineering (disjointed paths, specific routing such as delay minimization, optimization of flow routing according to network capacities); improvements in network operation in terms of monitoring and simplifications.