One Of Clear Linux's Kernel Patches To Help With Boot Time Proposed For Upstreaming

[phoronix]

Phoronix: One Of Clear Linux's Kernel Patches To Help With Boot Time Proposed For Upstreaming

Besides Clear Linux delivering often leading x86_64 Linux performance at run-time, when it comes to boot performance it has also been at the forefront -- in some configurations, can boot in 300 ms. Intel has invested significantly in ensuring Clear Linux boots as fast as possible for when running in the cloud or on containers in order to respond to increased demand as quickly as possible as well as for use-cases like Clear Linux within automobiles where they need to get automobile cameras active within two seconds of power on. One of their many kernel patches could be on its way to the mainline kernel...

http://www.phoronix.com/scan.php?page=news_item&px=Clear-Linux-ATA-Gfx-Boot-Up

[bug77]

Oh, cool. I need to dig up some of my old ATA drives and put them back in the case to take advantage of this

[andyprough]

If you want faster boot times just get rid of the fat. Use pure openRC or runit or S6 with gentoo or Artix or Void. systemd and sysvinit are where your fat lies.

[ssokolow]

If you want faster boot times just get rid of the fat. Use pure openRC or runit or S6 with gentoo or Artix or Void. systemd and sysvinit are where your fat lies.

Given that this optimization has to do with ATA init, it's about removing unnecessary serialization that occurs before initialization of and regardless of your choice for the init system.

What you said is akin to saying "If you want faster boot, ditch systemd or sysvinit" on an article about making BIOS or UEFI POST faster.

[FireBurn]

If you want faster boot times just get rid of the fat. Use pure openRC or runit or S6 with gentoo or Artix or Void. systemd and sysvinit are where your fat lies.

Pretty sure OpenRC still used sysvinit under the covers, also systemd seems to be much faster here on my Gentoo system

[Britoid]

Pretty sure OpenRC still used sysvinit under the covers, also systemd seems to be much faster here on my Gentoo system

The design of systemd does make it faster, it doesn't have to rely on a shell interpreter or user space binaries and can start things in parralel with a dependency graph.

However, modern distros these days tend to ship units configured in a way that makes it feel slow.

e.g. on Fedora, gdm.service requires network.target to be reached, however network.target requires firewalld.service and firewalld can take a couple of seconds to start.

gdm requiring network online is only really needed for network login, so a way of speeding this up would only require networking only when network login is being used, which is something Windows does already.

[andyprough]

Pretty sure OpenRC still used sysvinit under the covers, also systemd seems to be much faster here on my Gentoo system

Some systems, such as Devuan, use openRC with sysvinit scripts and initab instead of openrc-init. Artix uses openrc-init and openrc init scripts in /etc/init.d.

[andyprough]

Given that this optimization has to do with ATA init, it's about cutting removing unnecessary serialization that occurs before initialization of and regardless of your choice for the init system.

What you said is akin to saying "If you want faster boot, ditch systemd or sysvinit" on an article about making BIOS or UEFI POST faster.

"if you want your car to drive faster, don't throw the thousand pounds of weight out of the trunk - get more expensive tires!"

[bregma]

"if you want your car to drive faster, don't throw the thousand pounds of weight out of the trunk - get more expensive tires!"

If Monster Cable made tires....

[geearf]

Oh, cool. I need to dig up some of my old ATA drives and put them back in the case to take advantage of this

Isn't this done for all ATA drives, including SATA?