Linux is Greener, But Making Servers Sip Power Sucks


Chalk up another one for Linux. The open-source software was just deemed by Network World as greener than Windows Server 2008 when running as the operating system for servers. The computing magazine found that servers using Red Hat Enterprise Linux ran 12 percent more efficiently than those running Windows Server 2008.

But while that’s the headline conclusion (which, it should be noted, included many caveats), the real conclusion of the study is that it’s take meticulous attention and is difficult to get servers to run green. Primary power savings for the test came from the operating system asking the server CPU to dial back its performance. However, only chips made in the last three years even have the ability to “relax” when not being asked to perform computations. To add to the complexity, the software governing the hardware and chips themselves needs to be updated to allow for and manage a chip dialing back its cycles.

So beneath the OS, both the chip itself and the firmware need to have the most energy-efficient technology on board in order to achieve the maximum power savings with Linux. The power management features in the OS must also be turned on regardless of whether you’re using Linux or Windows. That needs to happen across all the servers in a data center to maximize power savings. And it’s possible that on some types of hardware Windows might perform a little better than Linux. One way or another, optimizing your server infrastructure for power conservation is as painstaking as building a house of cards.



That’s so true.

While you can actually tweak a system to make it run cooler, sometimes you just can’t, either because the hardware is not there yet or, in many cases, the kernel is not good enough (mostly for the last generation hardware).

And don’t even get me started on the desktop world. It’s even worse. There’s a reason why OEMs put Windows on notebooks.

Stacey Higginbotham

JC, not sure what you’re asking? The Network World Article is pretty in-depth on what was measured and how. Perhaps that will help?

Comments are closed.