Yes, it can be done by BGP. What was the question anyway?

Path hiding on a route server running BGP can happen if the route server has been configured to filter the chosen best path from reaching a particular route server client.
Assuming a traditional best path selection, when the same prefix is advertised to a route server from multiple route server clients, the route server will select one and only best path for propagation to all connected clients. Filtering that prefix out for a particular client, although alternate paths received by the route server might be appropriate, is referred to as “path hiding”.
The use of route servers (route-reflectors in general) has significant effect on redundancy by hiding alternate paths therefore a mechanism is needed which will allow propagation of multiple alternate paths for a certain IP-prefix.

What can we do?

A simple way to distribute multiple prefix paths from a route server is to use multiple separate BGP sessions for each different path. Well, this is definitely not scalable and it is undesirable for an IXP (Internet Exchange Point) where the number of BGP sessions is typically as small as two for each IXP member – one with primary instance of the Route Server, second with the backup one. So, forget it!

A much more clever approach was proposed to solve the problem – a BGP speaker could be allowed to forward multiple paths for the same prefix on a single BGP session. Another BGP capability (ADD-PATH) is used and negotiated bidirectionally between the route server and a route server client. Only the route server is operating with ADD-PATH “in mind” (that is in send-only mode) and clients are still sending the best paths only, but willing to receive more than one.

The ADD-PATH capability will find its way to many Internet Exchanges. We’ve seen implementation already in Junos OS, Cisco IOS XR, Bird etc.


References:
Advertisement of Multiple Paths in BGP
Internet Exchange Route Server, IDR Working Group draft
Internet Exchange Route Server Operations, GROW Working Group draft
Understanding BGP Convergence, Posted by Petr Lapukhov

Advertisements