FirstNight.org Statistics: 2000

December 2000 (First Night 2001) marked the 25th year of the First Night festival and the fourth year of our involvement with the site. The event web site at www.firstnight.org offered database-driven access to the entire schedule of events, a personalized planner feature, artist profiles, maps and other event information. The site receives very little traffic throughout the year but enjoys a massive spike immediately before and during the event.

The Platform

Firstnight.org, being a non-profit web site, can not afford the level of hosting that its two-day traffic spike would warrent. The best that could be done was find a moderately priced host that could, when needed, handle the traffic spike. We also required ColdFusion, uncapped monthly bandwidth, POP email capabilities, and certain other features. We settled on Communitech.net which offered all the features we needed for a price we could afford.

The site was on a shared, virtual IIS 5.0 server (Windows 2000) along with many other sites. The web server was a Dual Pentium III 500 Compaq Proliant configured with one gigabyte of RAM and fast SCSI disks. Bandwidth was not a concern as this box could not possibly saturate the available pipe. The site was written using ColdFusion 4.01. The database, also shared, was MS SQL Server 7.0 and hosted on identical hardware to the web server.

Year 2000 Statistics

Month of December (581kb Zip file of a 1,845kb Microsoft Word file).
This file was auto-generated by WebTrends.

Over 50% of the months traffic came after Christmas with nearly 24% of the total on the December 31st. The busiest hour was 1pm to 2pm on the 31st with 558,875 (9% of the month's total) hits. That translates 155 hits a second on average (hits represents HTTP requests, not "pages"). We served the most actual pages between 11am and noon with 8,419 page views. That's 2.3 pages a second. Of those about 43% represented database driven content pages.

The end of the browser wars was apparent: Internet Explorer made up 81% of traffic (80.82% version 5 and 15.67% version 4) and Netscape only 15.7% (96.19% version 4). All other browsers represented less than 3%.

On a sad note we had 18,342 internal server errors (response 500). That's 3.6% of total hits. Considering our circumstances however it wasn't nearly as bad as it may have been. A dedicated server or, better yet, server cluster would have drastically reduced, or perhaps even eliminated those errors.

Over the course of the month we transferred nearly 17 gigabytes of data. Embarrassingly enough there was also a huge number of 404 errors, mostly because a single-pixel spacer GIF image used was forgotten, accounting for over half-a-million missing file errors.