This will be more of a real BOF than a presentation.
Last year, the RA RSs public funding was terminated and now the RSs are being funding by the IXP operators.
MAE-EAST has been converted.
Sprint decommissioned the RA RS and is no longer making use of a RS supported by Merit.
DEC PAIX has a RSNG server.
Conversion at the various exchanges have been different for each one.
All the statistics available in RA RSs are/will be available in RSNG.
An on-line overview is available.
There was some talk at the last NANOG to use the route servers to support MPLAs using communities.
This is not available in the current configuration tools.
Is there really interest in this? Yes, there is interest. Some folks are doing this via PCH at MAE-WEST.
PCH is a local exchange.
For example, since PAC*BELL does setup the PVCs to facilitate implementation of MLPA, could they not set up the community objects in the RADB. Some folks think that is a bad idea because it confuses the level2/level3 involvement of the IXP operator.
There is also the level2 reachability problem. The core problem is that it is hard to reliably detect a remote unreachability.
Kevin Oberman: The RS should not peer with routers that don’t permit level2 reachability checks.
Stan Barber suggests that the status could be determined by asking the switches (ATM, Ethernet or
GIGAswitch) if the IX operators could make that information available to the RSs. Another person suggested that modifications of BGP could be a solution as well.
If the RS determines that there is a level2 reachability problem, it does not terminate a BGP session, but just withdraws the routes for that AS.
Abha notes that RSNG does include a NOC operation that can double check on any level2 problems.
There are three classes are users.
One uses the RS exclusively. Another peers with the RS for whatever information is available. There are groups that peer just for statistics purposes. Of course, there is non-class of non-users.
Send the peering template to rs-peer@ra.net. Also send suggestions to that address.
Who should administrate the objects related to MLPAs? Another person suggested that the community should be defined and folks who don’t want to peer with that community would filter against it and those who do would accept it. Those who are part of that community would insert that community into their relevant objects.
This would be like an AS-MACRO except that users could only add or delete themselves.
There is an issue on how MLPA might intersect with this.