Optimizing the Linux Kernel

Speed Test

Article from Issue 250/2021
Author(s):

We explore some optimizations designed to deliver a smoother experience for desktop users.

The Linux kernel is the core part of all GNU/Linux operating systems. The kernel is designed to run on a large variety of hardware, from web servers to routers and embedded devices. The default versions of the Linux kernel that arrive with the mainstream Linux distros are optimized for some very basic use cases. For instance, Ubuntu comes in Server, Desktop, IoT, and Cloud editions – each with basic optimizations tailored for the usage scenario.

Most distros make some effort to customize the kernel for its intended purpose; however, no one but you knows exactly how you are using your own system. You can tweak the Linux kernel in hundreds of different ways to improve performance or reduce latency. I'll outline some of those techniques in this article. Of course, some of these tweaks might have already been enabled by your distro's vendor; others are more specific and are seldom used at all. The goal of this discussion is to take you down inside the kernel and to demonstrate various performance-related optimizations. Needless to say, tricks with the kernel have the potential to destabilize your system. These ideas are best explored with a test system – at least at first, until you are sure everything is working.

I'll discuss a range of Linux kernel optimizations with the goal of improving perceived desktop performance, including smoothness and snappiness. Such things may have little effect in synthetic tests (such as the ones often conducted by Phoronix), but they can have a strong effect on the user. I am aiming this discussion at desktop and laptop users, including the significant number of people who need to run Linux on low-performance and legacy hardware.

Tinkering with the Current Kernel

You don't need to recompile the kernel to improve kernel performance. The easiest way to tweak the Linux kernel is to use the optional boot parameters that run at the command line when the kernel boots up. You can make temporary changes in the GRUB 2 boot menu by editing the line that starts with linux, or you can make persistent changes by changing the GRUB 2 configuration. The steps for changing the GRUB 2 configuration vary across different Linux distros. Sometimes you can use graphical GRUB 2 configuration tools (Figure 1), like the one shipped with YaST (SLE, openSUSE), or kcm-grub2, which is designed for KDE Plasma, or the command-line grubby utility. All of these tools change the contents of the grub.cfg file and then update the GRUB 2 configuration:

$ sudo grub2-mkconfig -o /path/to/grub.cfg
Figure 1: Editing the bootloader configuration is easy thanks to such helpful and friendly graphical front ends (in this case, kcm-grub2).

The following sections describe a few of the most useful kernel parameters.

elevator=

The elevator setting defines the Linux kernel behavior when distributing the I/O load on block devices (storage drives). This parameter defines the scheduler that will manage the I/O. Linux supports several different schedulers, which all have different strategies for balancing disk throughput and read/write latencies. Find out what schedulers are available and which one is currently used in your system with the following command:

$ cat /sys/block/sda/queue/scheduler

See the "Schedulers in Linux" box for more information on the available schedulers. The kyber scheduler is reported to have the best performance with mechanical hard drives (elevator=kyber), but if you are using a modern SSD or NVMe drive, it might be better use none to reduce the CPU overhead (elevator=none).

Schedulers in Linux

Modern hardware, with its enhanced support for multithreading and multiple CPUs, requires a new approach to scheduling. Linux is currently undergoing a transition to a new generation of multiqueuing schedulers. Consequently, some of the old schedulers that were popular in the past are gradually becoming deprecated. Ubuntu [1], for instance, has enabled multiqueue I/O scheduling by default in Ubuntu 19.10 onward and supports the following schedulers:

  • bfq (Budget Fair Queueing) – optimized for interactive response, especially with slow I/O devices
  • kyber – a simple algorithm supporting both synchronous and asynchronous requests; intended for "fast multiqueue devices"
  • none – does not reorder requests, thus consuming minimal overhead (multiqueue version of the old noop scheduler)
  • mq-deadline – multiqueue version of the old deadline scheduler

Older schedulers, such as cfq (Completely Fair Queueing), deadline, and noop are deprecated in recent versions of Ubuntu and other distros; however, they are still used with older systems and in situations that do not require multiqueuing.

Staggered Spin-Up Elimination

Even if Linux is installed on a super-fast SSD, the boot process can get slow if any other rotational hard drive is attached to the computer. This issue is called staggered spin-up, which means that the OS probes ATA interfaces serially, one by one, to reduce the peak power consumption. Normally, desktop users do not benefit from this default configuration, and they often feel annoyed with the longer boot times. To see if your system is using staggered spin-up, enter:

# dmesg | grep SSS

If it is, eliminate the issue by passing the following boot parameter:

libahci.ignore_sss=1

Buy this article as PDF

Express-Checkout as PDF
Price $2.95
(incl. VAT)

Buy Linux Magazine

SINGLE ISSUES
 
SUBSCRIPTIONS
 
TABLET & SMARTPHONE APPS
Get it on Google Play

US / Canada

Get it on Google Play

UK / Australia

Related content

  • Working with the Kernel

    If you work with third-party hardware drivers, or even if you just need to fix a broken system, someday you might need to upgrade the Linux kernel.

  • Performance Tweaks

    If you are looking for ways to speed up your Linux, consider this collection of curated performance tweaks.

  • Kernel Tips

    Worried about a recent security exploit? Want to take advantage of a new hardware feature? You don’t need to be a Linux expert to patch and compile the Linux kernel. We'll show you how to get started.

  • Compiling the Kernel

    While not a requirement, compiling the Linux kernel lets you add or remove features depending on your specific needs and possibly make your kernel more efficient.

  • Nitrux 1.5 Ships with Kernel 5.13

    Debian-based Nitrux Linux is the first distribution to ship with kernel 5.13

comments powered by Disqus

Direct Download

Read full article as PDF:

Price $2.95

News