Watching Dearborn

Got up in the dark in Los Altos this morning to watch and listen to the technical sessions at Nanog 47, taking place in Dearborn, Michigan. The quality of the audio was excellent; the presenters' slides were available as pdf, and the video of the slides was good enough that one could tell which slide the presenter was using. During the question and answer sessions the camera views included parts of the audience; there were many faces familiar from previous meetings, including the Nanog in Santa Domingo in January, and the one in Philadelphia in June.

It seemed like one had the experience of less than half the meeting - no hallway conversations, no booth setup duties, no working out the detailed logistics of the hotel. In strong positive balance, I drank my own tea, and slept in my own bed, avoiding cars, airports, and flights. Maybe one out of three or two out of three meetings a year is a good balance.

Content summary, 19 October 2009

The Internet is still growing, and the top sources/sinks of interdomain traffic have changed significantly in the last 2 years. Only 150 origin ASNs now contribute 50% of traffic (compared to 1000s two years ago). Arbor are running the ATLAS observatory, described here. Jim Cowie from Renesys came separately to the same conclusion - he'd expected to see a dip in traffic, as they did in 2001, but is only seeing, perhaps, a slowing in the growth rate.

ARIN is making policy about address assignments in the IPv4 and IPv6 space which have consequences for routing policy; in particular, the current policy about aggregation makes traffic engineering difficult. There was an eloquent comment from the floor about how to avoid creating the same swamp in IPv6 addressing as has happened in IPv4 addressing.

Matthew Petach took great notes, in the morning, and afternoon.

After the main sessions there was a peering BoF, at which Hurricane Electric baked a cake for Cogent - Yahoo photograph ,here

Tuesday 20 October 2009

Greg Hankins from Switch and Data (as was - Equinix subsequently announced an acquisition) reviewed the 40G and 100G state of standards and hardware.
Good presentation, with pictures. here

Microsoft Research presented intra data center traffic pattern measurements, suggested using VL2, VLB. "Developers want network virtualization; want a model where all their servers, and only their servers are plugged into an ethernet switch" See here
AMS-IX have gone to VPLS/MPLS and a much more complex design for the exchange - they didn't think they could continue to scale up with the Brocade switches which had the largest number of 10GE ports on one switch if they continued doing L2 only. AMS-IX
More Matthew Petach notes - morning , and afternoon .


had more IPv6 deployment stories - Matthew had described the Yahoo experience on Tuesday as a lightning talk. All the talks are on the agenda here

Historical note - last time I was in Dearborn was in late 1988 or early 1989; Ford had just decided that they would not install an additional supercomputer for solid modelling design in Europe, and instead would pay for a transatlantic TCP/IP connection between Ford Dagenham (in SE England) and Ford Dearborn. Richard Troiano, cisco system's first sales guy, went with me to meet Scott Anderson. Ford HQ is just across the street from the Nanog hotel.