Short bio: Computer Scientist, FOSS supporter (read more)
Tux Machines (TM)-specific
For years, the Blartner Group and various other prestigious research firms have issued benchmarks claiming that Linux is inferior to Windows. Naturally, we at Humorix have always been skeptical. But while scraping the bottom of the barrel for story ideas last week, we decided to see if we could duplicate any of these benchmarks on our own.
It turned out to be a lot harder than we thought, but we finally concocted a scenario in which Windows outshines Linux.
First, the Humorix Vast Research Center & Basement of Doom(tm) wanted to try a fair, head-to-head matchup between two computers with identical specs. Obviously, a leading-edge company such as Blartner would immediately skip this worthless step in one of their benchmarks, but we wanted to cover all of our bases.
Using every measurement tool we could find, the machine running the latest version of Red Hat and Apache simply outperformed the machine running the latest service pack of Windows and IIS. Even after tweaking the settings to handicap Linux -- such as downgrading to an unstable 2.1 kernel -- we still couldn't make Windows look good.
Next, we installed our secret weapon on the Linux box: killrandom(3). Every few seconds, this daemon randomly picks a victim process to kill, thereby simulating the Windows experience. Even with killrandom running at full blast, the Linux machine still achieved respectable performance.
Frustrated by our lack of success, we finally moved to the "nuclear option": setting the number of allowed IIS and Apache users to zero. We then performed a simulation of the Slashdot Effect on both machines. The Windows machine was able to serve an amazing 2179 "Server Too Busy" error messages in one second. Meanwhile, Linux and Apache only managed to generate a meager 1673 such error pages.
"Windows is very good at doing nothing fast," joked one of the Humorix lab assistants during the benchmark study/keg party.