A couple weeks ago, a customer of us asked: Can you tell me if I can run an SR-PCE on Red Hat OpenShift?

The answer from us was a resounding: Yes, absolutely you can !!!

And thus started the exciting journey to create a mini Segment Routing topology, along with an SR-PCE instance to prove to the customer that you can indeed run an instance of a containerized router and use it as an SR-PCE for your Segment Routing Infrastructure.

What is an SR-PCE?

Simply put, as we talked about in our book, a Segement Routing Path Computation Element (SR-PCE), is type of an SDN controller used by Segment Routing Traffic Engineering (SRTE).
In general, a PCE is meant to perform the following three functions:

• Topology collection so that it has a full view of the network topology. This could be done through BGP-LS.
• Path Computation Algorithm so that it can create a path for the desired traffic across the network according to defined contrainsts.
• Communication with Headend so provide optimal path, than can be dynamically adjusted based on realtime network usage. This communication is done via PCEP - PCE Communication Protocol.

For the purpose of this blog, we used Cisco XRd to be used as an SR-PCE. In our previous blog we talked about what is Cisco XRd and how to run XRd on Red Hat OpenShift. Check it out if you are new to Cisco XRD.

Segment Routing Topology on OpenShift

The following topology comprising 7 XRd instances is created when you succesfully executing the code in this public Git reposity. Notice the use of Linux Bridge for point-to-point link between routers, which are mapped to IOS-XR interfaces.connectivity each router.

Note: This network is fully contained within a Singe Node Openshift (SNO) cluster. It does not implement, nor require, any external connectivity

Upon successfull execution of the code in the repo, the included configmaps will pre-configure the XRd instances to create a fully functional IGP and BGP Topology, along with the PCE-PCC peering session using PCEP as shown in the topology:

Notice the lack of route leaking between the ISIS L1 and L2 domain and the BGP-LU peering between the nodes. The topology create here is based on the RFC-3107 Seamless MPLS architecture that is fouind in majority of modern large-scale deployments.
The rest of this blog out lines how to run this topology using the code provided in the Git Repo at https://github.com/k1slam/public-xrd/tree/main/xrd-sr-pce-sno