I was on vacation in Hawaii on the big island in probably `94 or before and fixed the network at the Volcano Arts Center. The director gave me a nice discount for anything at the center and we ended up going to a mutual friend's house, where they fixed us a traditional Hawaiian dinner. I also had to hack my Pentax MX camera when the shutter button fell off while we were on top of Puu Oo looking down into the most breathtaking view I've seen.

Awesome vacation. And, no, not unusual, for this crowd.

Chris.

I spent much of last Friday (before the IETF meeting started) tying the hotel wireless infrastructure to our dual gig uplinks in the convention center. An amusing part of this was that to find the fiber that went between the convention center and the hotel required us to dig out our own documentation from when we used this facility in 2005. One page allowed us to locate where the fiber terminated in the conference center--a room on the roof of the center. We then had to get one of the conference staff to climb into the equipment cage as they never used it and had lost the key.

Imagine using an elevator that required calling security to enable it to go to the top floor, winding around and ducking under giant HVAC equipment in a dust-covered room, and turning on the lights to see a 8-foot high equipment cage. Imagine the staff member using a pallet as a stepping stone to get into said cage. Imagine that the lights are on a timer and go off automatically in 5 minutes, leaving said staff member to try to climb out in the dark. Fortunately, I was there, outside the cage.

It's always interesting when we know things about the facility that the staff doesn't.

That said, both the convention center and the hotel staff have been great to work with, the key has been found--and no one has been fired, as far as I know. Certainly the same folks I've been dealing with are still employed here.

Chris.

We tried. We sent in one of our best:

http://cavebear.com/index.php?option=com_bookmarks&Itemid=67&mode=2&catid=-1&navstart=0
Search for ICANN.

Chris.

Please read up on what the IETF does. We meet to design the protocols that allow the Internet to work. Our primary focus is not to meet people or learn new things, although both things happen as secondary outcomes. Your ability to transmit your message above is because of the work we do.

The selection of the sites for the IETF meetings is complex and many variables go into the mix. We try to encourage participation in the IETF around the world while somewhat equalizing the inconvenience of travel by having meetings at locations around the world. Few cities have conference centers with enough meeting rooms in the right mix of sizes for us. We are very sensitive to price, while wanting the facility to allow us to install our own network. There has to be enough hotel rooms close to the conference center for 1200-1600 attendees. Adequate restaurants, climate, stability of the government, openness of the Internet in the country, and an almost endless number of other variables are used. Paris was chosen for many reasons.

We had two 1G uplinks. No, (almost) no participant has better internet connections at home.

Chris.

We do. However, most of the time that gear is used for the convention center and not for the hotel. This time we deployed 50 Cisco 1200 APs, 11 24-48-port switches, 9 smaller 8-port switches (mainly for rooms needing multiple wired connections), and two Juniper routers, each connected to a 1Gbps uplink using different paths to the rest of the Internet.

We did consider deploying some of our leftover APs (some 20 more 1200s, and some 1131s that we only use for outlying areas) in the hotel, but there really wasn't time, nor was it practical due to the number of floors (33) and the number of access points already on each floor (10). In Taipei we deployed around 20 APs in the main hotel. Each venue is unique.

Chris.

The solution was in many changes. That included changing the lowest allowed speed and broadcast/multicast rate to 2Mhz (from 1) for the 2.4G APs and changing the receiver sensitivity on the AP radios by changing the Colubris wireless interface config "distance" from "large" to "small". Often it is a set of small changes that, together, effect change far larger than they do individually. You can see the incremental improvements in the bandwidth graph in the presentation.

I suspect that the channel reallocation was the most important change, but the time is too short to test all combinations and determine the impact of each change.

Chris.

Well, to be pedantic, 8, actually. I used 36, 40, 44, and 48, two on each end of each floor (each floor of the concorde looks like a 2D football).
Chris.

We will remove the connection to their uplink, so performance will go down when we leave. However, they are keeping our configuration changes, so I expect the performance to still be significantly better than it was before we arrived.
Chris.