The Load Scheme

        In the document root of the web server, I placed 5 HTML files and 1 XML file (with no images), that average about 2kb each (smallest is 400bytes, largest if 8kb).  I configured WebLOAD to request each one of these files in order to complete a "round."  I setup the load generator to step up the load from 1 concurant user to 40 concurant users, adding a user every 20 seconds, with each user just requesting round after round with no delay.  I setup "vmstat" to record resource usage to a file every 20 seconds durring the test. 

        To be fair, I ran the test 3 times for each server (lighttpd and apache), and then took the middle result (as a complete set).  To prevent rampup and other high load issues, I ran each test twice.  Once to load the server, and the second time to capture the data.  So in total, there were 12 runs made.  I compared result sets, and picked the one that was in the middle (the others we not off more than one percent).