Within the results page, we track the power performance of the Linux mainline kernel, and the effect of various power saving features under development for sample mobile, server and desktop platforms. This helps the Linux community to track the evolution of power performance of the mainline kernel under development.

The 2.6.22 kernel is selected as the starting point of our measurement and power change relative to this kernel is displayed in the graphs. Currently the i386 version of the kernel is used as tickless idle has not been merged into the x86-64 architecture in the kernel.org kernels, but you can get the patch for it here. Hopefully this patch will go into the 2.6.23 kernel.

Test Systems

• Server - a dual-socket 2.13 GHz Quad-Core Intel® Xeon® CPU with 4 Gb RAM
• Mobile - a 2 GHz Intel® Core™ 2 Duo CPU laptop with Intel 965 Express GM integrated graphics, 1 Gb RAM
• Desktop - a 2.667 GHz Intel® Core™ 2 Duo CPU with 2 Gb RAM

Workloads

• Mobile, Desktop - The battery life tool kit (BLTK) workloads that simulate the following: idle condition, office document edit, a kernel developer editing and compiling kernel files, reading a document via web browser, and playing a DVD movie.
• Server - A webserver workload is applied. Http requests are sent to the server at a constant rate with the httperf program from another machine via a cross-over cable. Each web request is handled by Apache with a lookup in MySQL database.