Want to optimize 100G/400G link performance? Try Segment Routing
Does news of the High-Luminosity Large Hadron Collider launch in 2029 strike fear in your heart? Maybe, if you are working with a Research Education Networks (REN) that will be in receipt of the near five-times increase in datasets once CERN ups the number of collisions it needs to study fundamental components of matter.
And that’s just CERN. There are many other data-intensive projects whose many petabytes of data will need to transit the global network of RENs quickly and securely.
While many RENs are adding 400G links to handle this onslaught of traffic, these backbone upgrades won’t happen all at once. And some network operators may choose to keep both 100G and 400G links. This mélange of capacity and speeds at different points in the network may have some operators asking how best to optimize utilization of their IP networks.
It’s worth considering a fresh approach like Segment Routing (SR)
SR establishes predefined forwarding paths in the IP network that override the default shortest path while meeting specific constraints like available bandwidth, latency and physical diversity. This option removes the need for resource-intensive, per-path control plane signaling and keeps a minimum of forwarding state information on routing nodes. As a result, it scales better than conventional approaches based on RSVP-traffic engineering (RSVP-TE) or label distribution protocol (LDP) and can engineer more granular traffic flows.
While LDP is good at finding the shortest path for packets to travel, its shortcoming is that it can only pick one path. And as any driver knows, if there’s only one highway where you need to go, it’s going to get congested.
Segment Routing will spread the traffic out so that the packets can travel over different links – determined by the intelligence imbedded in the edge node. It will also know if it’s a 100G or 400G link – and subsequently distribute four times more traffic toward that link, providing a more balanced utilization of the network. And if any link should fail, only a portion of the traffic is impacted rather than all of it in the case of LDP.
Moreover, SR can be applied to pre-calculate backup paths that protect against failure of the primary path. Primary and secondary paths can be defined with strict or loose hops and initiated by a router or by an external path computation element (PCE) controller. Fast reroute is supported through the ability to precompute alternate next-hop backup paths called loop-free alternates (LFAs), to quickly take over if a failure is detected.
Because Segment Routing can take advantage of the new higher-capacity links as you gradually add more capacity to the network, you also don’t fall into the RSVP trap – in which you set up a traffic engineered tunnel and must stick with it.
Instead, Segment Routing assesses the traffic engineering parameters you have established for the links and finds the best path.
Segment Routing takes a source-based routing approach, which requires only the ingress or head-end router to maintain policy and state information about the path. In its simplest form, a segment route is a sequence of segments that act as waypoints for forwarding packets along a constrained path that meets a given policy. Segments can refer to different network objects such as subnets, routing nodes or interfaces to be included or excluded in the data path.
Add in a centralized software defined networking (SDN) controller and you can tap into better visibility of the whole network. Not only the capacity of the links but also how much of the link is being utilized at that moment. Segment routes can be engineered and programmed by the controller or determined by edge routers themselves based on policy constraints.
Because SR-MPLS supports native IPv4 and IPv6 data planes without adding packet processing requirements, it can be deployed on existing hardware and yield immediate service benefits – without sacrificing any transport capabilities offered by LDP and RSVP-TE. Using the SR toolkit, operators can incrementally add the features they need to an existing LDP/RSVP network with the options they are comfortable with. This approach eases migration and allows the operator to gain operational experience before introducing more powerful features.
With a decade of history behind it, Segment Routing has shown that it provides more effective tools than RSVP-TE and LDP to reliably control and steer IP traffic through the network along optimal routes. And from a user perspective, SR-TE LSPs are similar to traditional, traffic-engineered label switched paths (LSPs) such as RSVP-TE and offer a natural migration path.
SR is a powerful and proven technology for deploying scalable and programmable IP services that meet deterministic service level objectives for cost, performance and reliability. And as backbone networks begin the transition from 100G to 400G links, SR offers a solid alternative for network operators wondering how best to optimize utilization of IP networks – avoiding points of congestion and ensuring all available links are utilized.
Interested in finding out more? Check out our Research and Education Networks webpage, the open source Containerlab initiative and join our community on Discord to talk to peers and Nokia subject matter experts.