katrinacheckin.org fallout

What a strange 24 hours.
As the last post says, I launched KatrinaCheckIn.org yesterday morning, around 2:30 a.m. By the time I woke up, around 8:00, I had CNN on the phone asking me to do an on-air interview the next morning. I also had visitors pouring in at the at-the-time-unbelievable rate of 1,500 an hour.
People were posting about their conditions and their loved ones — heartwrenching stuff in many cases. I was happy, because the site was fulfilling its purpose. There were a few other similiar sites set up at about the same time, but I think that KatrinaCheckIn.org was the largest of the independent ones. (Sites set up by the Times-Picayune and WWL-TV were probably the same size or a bit bigger.)
Anyway, I was frustrated when the site started crashing around 9:30 a.m. I wrote to my web host — which, for the record, is not the crappy one I recently tossed to the curb — to complain. The server ended up fixed, but it failed again a couple hours later. When the site was up, it was grindingly slow. I was a little mad.
Which was dumb. Because, as it turns out, I was the one crashing the server. Because KatrinaCheckIn.org was drawing in an absolutely insane number of hits.
In the end, my host — completely reasonably — had to take the site down. Turns out the site was on pace to pull in 84 million hits in a single day. (To put that in context, here’s a note from my host: “Companies like realtor.com have 60 load-balanced web servers, a large staff (and they spend 40 million a month) to do 20 million a day without issues. Yahoo’s Sept 11th memorial 3 day ‘event’ did 64 million hits over 3 days, took months to plan, was done on 500 servers and was fully staffed. That’s the kind of traffic that showed up here.”)
The wholesale bandwidth alone for a day’s traffic on KatrinaCheckIn.org was going to be $18,000. The site was generating 1,000 requests to the web server per second right before being shut down. (Again, from my host, “Wikipedia for example does a 1,000 requests/second on a 64-server cluster with about a dozen of those being squid proxy caches.” I don’t know what a squid proxy cache is — although it sounds totally awesome — but Wikipedia’s 64 servers compare quite favorably to the one server KatrinaCheckIn.org was on. And it shared that server with hundreds of other sites — crabwalk.com among them.)
So — what now? The terrific people at Textdrive are working on a few possible routes to bringing it back without melting servers and emptying my 401(k). May have word soon. I’m hopeful. But I wanted those who used the site to know I’m sorry it disappeared, and that I’m working to bring it back.
(FYI, CNN called late last night and cancelled the interview — presumably because they realized there was no site left to talk about.)