Date: Tue, 20 Feb 2001 23:41:58 +0800 From: Adrian Chadd <adrian@creative.net.au> Sender: owner-nanog@merit.edu
On the same breath, why aren't people using route-servers at NAPs?
There are probably three reasons: 1. People don't understand the concept of the route servers. They know about setting up a traditional peering, though. 2. People worry (justifiably) of black-holing traffic because the route server assumes that all of it's peers have the same reachability to it as they do to each other. This is normally the case, but there are several cases of failures at NAPS where this was not the case. Al were transitory, but they did result in loss of traffic. At least One major provider refuses to use RSes for this reason. 3. People do not maintain their routing policies in the IRR causing the route servers to not advertise routes that they should. I spoke with the NOC of a "tier-1" which used the RAs at NAPs but had not updated it's policy in YEARS! They had picked up a few customers since then and the route servers were advertising less than 10% of the routes the provider was carrying. Makes a RS peering pretty useless. nd there have actually been many such cases. Take a look at http://www.rsng.net/rs-views/. Pick a route server and then a participant with a non-zero number of peers exported to. Check the number of "grey" routes vs. the number of "green" routes. Grey routes are those not in the provider's policy. Green ones are. For MANY providers, the number of grey routes exceeds the number of green route. This makes use of the route servers at least a bit problematic. We have switched several peers to direct peering for this last reason. R. Kevin Oberman, Network Engineer Energy Sciences Network (ESnet) Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab) E-mail: oberman@es.net Phone: +1 510 486-8634