Open Source Subnet An independent Open Source community View more

Speed up Linux: No kernel patch required

Simple tweak gives major speed boost

Want to speed up your Linux desktop without compiling a new kernel? You don't need a 200-plus line patch for the Linux kernel when a couple of lines of Bash will do the trick.

A few days ago a kernel developer posted a patch to the Linux kernel that changes the way the Linux "scheduler" works. For non-geeks, this is the way that the kernel hands off tasks to the CPU. This has been a topic of a lot of debate over the years, with kernel developers proposing dueling schedulers and sometimes storming off when their proposal was rejected.

So there was a lot of buzz and excitement when the patch from Mike Galbraith, clocking in around 225 lines of code, showed a dramatic improvement in desktop latency. All is well and good that the patch works, but it would be a long time before most Linux users would see an update. It won't be for a few weeks before it makes it into the mainline kernel, and six or seven months before it trickles down to users. Some users are willing and able to recompile their kernel, or willing to install patched kernels from third-party sources, but most users don't fall into those categories.

Turns out, users don't have to wait if they're willing to make a few small modifications to their systems involving a few lines of Bash code added to a system configuration file (/etc/rc.local) and a user's login file (.bashrc). That comes from Red Hat's Lennart Poettering.

See the post on Web Upd8 for instructions on Fedora and Ubuntu machines. I've tried the second method on a machine running Linux Mint 10 (which is Ubuntu 10.10 based). In decidedly unscientific testing, it seems to produce an improvement in several areas — particularly when using Firefox or Chrome. I haven't yet tried the kernel patch yet, but according to Markus Trippelsdorf, the user-space changes reduce latency more.

The immediate effect may be a speedup on the desktop for Linux users based on Poettering or Galbraith's approach. But even better, maybe this will kick off a new round of competing ideas on speeding up the Linux desktop.

Editors' Picks
Join the discussion
Be the first to comment on this article. Our Commenting Policies