End to End Segment Routing

Segment Routing (SR) is a kind of source routing technology which means that the source can define the path that the packet will take. In source routing with SR, the hosts determine the path, encodes the path information to the packet header. Segment Routing divides a network into several segments and assign a segment ID to each segment, Prefix SID, Adj-SID etc. A Segment is and instruction that a node executes on the incoming packet. Format of a SID depends on the implementation; it may be an MPLS Label or IPv6 Address.

SR architecture doesn’t have a specific control plane protocol. A routing protocol, ISIS, OSPF or BGP, is generally used to distribute Segment ID information in the network. Segment Ids distributed by IGP are called IGP Segments and those distributed by BGP are called BGP Segments.

The Seamless MPLS adopts core, aggregation, and access networks separation with different MPLS/IP domains that are isolated from IGP perspective. Within a domain, both ISIS and OSPF can be a choice as IGP. Reachability across domains can be achieved by using RFC 3107 procedures whereby BGP-Labeled Unicast advertisements are exchanged between different domains, namely Core, aggregation and access.

  • In Inter-AS designs, the iBGP labeled unicast is used to build inter-domain LSPs inside the AS, and eBGP labeled unicast is used to extend the end-to-end LSP between ASs.

Inside the access ring you may run diffierent IGP domain. SR is enabled with the TI-LFA feature inside this domain. It is suggested to have different IGP domain in the all access nodes. And the core network will be different IGP domain as well.

  • Each of the domains, like core or access, is running independent IGP with SR. Between the different IGP domains only loopback address of Service RRs is imported to other domains. Aggregation routers loopback addresses are also imported into both domain which they have connection.
  • Fast convergence within each domain is achieved with Topology Independent Loop Free Alternate

(TI-LFA) technology using SR/IGP capabilities. Inside domain you can simply enable TI-LFA. It will automatically calculate backup paths for every destination and install them into forwarding table.

  • The IGP in each domain could be OSPF single area (preferably area 0) or ISIS single level (preferably level-2). Considering OSPF area or ISIS level-2 may be helpful for your future extension inside the domain.
  • Each access domain is connected to the core domain using two IGP Aggregation Area Border Routers .
  • End-to-end transport is achieved using BGP LU with BGP prefix SID.
  • Each access node and Aggregation node participate in BGP LU. (Red arrows show the BGP peering for labeled unicast.
  • Each Aggregation node has iBGP peering (nexthop-self) to two dedicated transport route reflectors (RR).
  • Each Aggregation node is an inline RR for all corresponding access nodes in its region.
  • For the service route advertisement, each service access node have iBGP peering with service route reflectors.
  • End to end protection BGP FRR can be used.
  • If the network is larger and has pre-aggregation layer, it is suggested to consider pre-aggregation layer and aggregation layers as different IGP domains.

Control Plane

Data Plane

Suggested Readings

About: fabricplane