The previous section’s explanation of the path selection algorithm described path selection logic based on properties of the two endpoint nodes and the nodes present in paths between them. It implicitly assumed that each node of the graph had this information available, but did not address the details of how it was obtained.
This section describes a BGP-based reference implementation that provides the mechanism for both distributing this data to the nodes of the graph/network as well as for executing that path-selection logic that acts on that data.
In this reference implementation, each node of the graph corresponds to a router running BGP, and each edge of the graph corresponds to a BGP session between two nodes.