The Slashdot Effect

The graphs below represent hits-vs-time and bytes-vs-time for July 26 2000, the day the AUUG/LinuxSA InstallFest story hit Slashdot. The surge of hits and bytes resulted several thousand visitors to the InstallFest photographs here, here, and here.

The normal "background radiation" level of hits to this web server is something in the order of 1000 - 1700 hits per day. Shortly after the story was posted it jumped up to over 2200 hits per hour. It still hasn't quite died back down to its normal level at the time of this posting, over 24 hours after the appearance of the story.

The graphs were produced with the perl GD::Graph::lines module, available on CPAN, coupled with a 30-line Apache log analysis script I wrote for the purpose of working out hard the Slashdot Effect would hit.

It should be noted that the links to this site were relatively obscure by Slashdot standards: Readers needed to click the Read More link, then the link to the LinuxSA website, then the link to this site, in order to hit my web server. It'd be reasonable to assume that higher-profile Slashdot stories would exert an even greater load on a typical web server.

This site is served by an Agile ethernet carrier service running at several megabits per second between the Ethernet switch at my home and a switchport in the computer room at Internode Professional Access, my ISP and employer. Agile Communications and Internode were both sponsors of the InstallFest event.

Mark Newton
newton@atdot.dotat.org