Timing The Boot Process From Linux 3.0 To Linux 3.6

[phoronix]

Phoronix: Timing The Boot Process From Linux 3.0 To Linux 3.6

Here are Bootchart results indicating the boot speed from the Linux 3.0 kernel through the latest Linux 3.6 development kernel...

http://www.phoronix.com/vr.php?view=MTE4NTI

[FireBurn]

Yet another pointless "article"

[89c51]

Michael could you do a systemd VS upstart.

[not.sure]

lolwhut?
There are still people who boot more than once or twice a month, and care about boot speeds? That sounds like 90ies.

[halfmanhalfamazing]

lolwhut?
There are still people who boot more than once or twice a month, and care about boot speeds? That sounds like 90ies.

I shut down constantly. I don't leave the computer running when I leave the house or sleep at night. So daily, I boot the computer at least once.

And apparently:

http://blog.martin-graesslin.com/blo...p-performance/

Login speed and overall computer boot speed are similar items. Who cares about it? A lot of people do.

And yes, I do include auto-login into the boot speed equation. Just because the computer will boot into the login screen in 8 seconds, that does not count. If it boots into usable desktop mode(be it XFCE, E, GNOME, KDE, etc) in 21 seconds, then the boot time is 21 seconds. Not 8 seconds. (these aren't my boot times, I'm just throwing out random numbers)

[werfu]

Seriously WTF? 11s to 48s is hell of huge delta.

[droidhacker]

lolwhut?
There are still people who boot more than once or twice a month, and care about boot speeds? That sounds like 90ies.

Laptop. Boot speeds actually *do* make a difference there.
Mine boots from power button pressed to gnome-shell in about 10 seconds.
Obviously, I don't leave it plugged in all the time, that would defeat the point of a laptop.
And standby empties the battery in about 3-4 days.

So yeah, a quick bootup is highly desirable.


HOWEVER,
This article is retarded. He's measuring a FULL bootup to judge the difference in the KERNEL??? I SAY NO WAY!!!!

[grotgrot]

There are still people who boot more than once or twice a month

Yes - I have a laptop. The hard drives use an ATA password which doesn't provide protection while the system is powered on (or in standy). Similarly I use encryption (LUKS/dmcrypt) and everything is unlocked after you power on.

Somewhat related, the time to resume from hibernate/suspend to disk is longer than a clean boot.

On my desktop I sometimes have to boot into Windows (about 3 times a week when I can't use virtualization). Ubuntu doesn't multiple swap partitions for hibernate so I have to go through a full restart cycle.

[oliver]

I let my laptop sleep constantly. It also doesn't use any binary blobs.

My desktop, while using the fglrx drive, suspends just fine. Yet I shutdown at night when going to sleep, and only boot it up after I get home from work. There's no need letting your PC idle all day (I used to do that when I didn't have to pay my own power bill).

So yes, boot times should still mater. Also, don't expect everybody to suspend in an office enviroment, just shutdown at the end of the day.

That said, when measuring kernel boot times, only measure the kernel, or maybe yes, both. kernel that init's into busybox for pure kernel time, kernel to console and kernel to logged in GUI. And then, which this article really really misses, a table with the results as a conclusion. I'm not going to open 6 or whatever bootchart windows in a browser to compare. A little table, 3.0: 8s; 3.1; 12s; etc is missing.

[Paradox Ethereal]

Os-jitter.

Michael, why not try and get into OS-jitter. Many-cpu machines benefit from low os-jitter, so big guys want it. On desktops it translates to low-latency, smooth framerates and responsive desktop. A benchmark on what kernel-components has the lowest os-jitter would be interesting.

Peace Be With You.