Routing WG - RIPE 64
Wednesday, April 18th, 14:00 – 15:30
WG co-Chairs: João Damas and Rob Evans
Scribe: Vesna Manojlovic
Archives of presentations & chats: https://ripe64.ripe.net/archives/#Wednesday
Stenographers transcripts: https://ripe64.ripe.net/archives/steno/14/
A. Administrivia (5 minutes)
- Welcome, apologies from co-Chair
- Thanks to scribe, jabber scribe and stenographer
- Microphone etiquette
- Minutes of RIPE 63 approved
B. The BGP DFZ in 2011
(30 minutes) - Geoff Huston, APNIC
The presentation is available at: https://ripe64.ripe.net/presentations/25-2012-04-16-bgp2011.pdf
Closing words: "[Growth of the routing table] is not the major problem out there in the eBGP world. We could do better. We should do better. But, what we're doing so far isn't exactly dangerous and it's certainly not killing your router."
Dave Freedman, Claranet, asked if Geoff measured the effect of the drive-by deaggregates.
There is a whole new industry of DDoS mitigation providers, and they are using the method of rapids of de-aggregation for "sinking" traffic.
Geoff answered that if they do it in bulk, he would expect to see a spike in the graph.
Carlos Martínez, LACNIC, asked if Geoff's conclusion that the eBGP is not growing much could be taken as an indication that the black market in IPv4 isn't happening, because it's hard to imagine how this could happen without deaggregation.
Geoff replied that as part of APNIC's transfer policy, they publish a list of transfers. This makes the market public, not black. He added that it's still early days and they'll have to see what happens when the RIPE NCC runs out. He added that it would seem that routing people are quite conservative and that routing systems have consistent numbers.
James Blessing, Limelight, asked (via Jabber) what prefixes have more-specifics as opposed to the "whole" prefixes, solid prefixes.
Geoff replied that classification is quite difficult.
James Blessing, Limelight, asked (via Jabber) : the IPv6 numbers on Geoff's slide stop below 30k when we are expecting to see N per ASN, N>1. When do you see that happening?
Geoff replied that: /48s are widely accepted; there is not a lot of deaggregation in the IPv6 routing table. There's no legacy swamp. He added that currently folks are not on the edge of the network, there is no traffic engineering juggling in IPv6.
Geoff also said: "Europeans in the routing land don't take holidays".
Randy Bush, IIJ, commented that there is a paper available (2010 paper in JSAC) that answers James' question.
C. Route Flap Damping Considered Usable
(20 minutes) - Randy Bush, IIJ
The presentation is available at: https://ripe64.ripe.net/presentations/136-120418.ripe-rfd.pdf
Andy Davidson, Hurricane Electric, asked if they could reduce the amplification effects of dampening the prefix by tweaking defaults, using dynamic dampening.
Geoff answered that there would be a way of doing this by attaching timers to every prefix but the problems would be re-coding, working with a delicate balancing, bumping RAM, etc.
Randy answered that his solution would be to change only two constants and one line of code. He added it was a horrible hack.
Joao asked if anyone was using open source implementations like Open BGP or Quagga to test this.
Geoff replied that they are using Quagga, and that it is possible to play with it to implement prefix timer idea.
D. BGP Route Stability
(30 minutes) - Alexander Asimov, HLL
The slides are available at: https://ripe64.ripe.net/presentations/154-ripe64.asimov.route_stability.v8.pdf
There were no questions.
Z. AOB & I/O With Other Working Groups