Uh, if you're comparing two quantities like that, requests per second of a test jpeg for 2 servers, your bar graphs REALLY ought to have 0 as the lower axis. Not 500, or 1600, or any other value. Doing it the way these graphs have been done is hugely misleading to anyone casually perusing the data. Sure, lighttpd is faster by these metrics. But not by as much as these graphs lead one to believe. (unless you're paying really close attention.)