Re: [gentoo-user] new machine : start-up time

["Canek Peláez Valdés" <caneko@gmail.com>]

On Sat, Oct 13, 2012 at 4:14 PM, Volker Armin Hemmann
<volkerarmin@googlemail.com> wrote:
> Am Samstag, 13. Oktober 2012, 15:57:31 schrieb Canek Peláez Valdés:
>> On Sat, Oct 13, 2012 at 3:40 PM, Canek Peláez Valdés <caneko@gmail.com>
> wrote:
>> > On Sat, Oct 13, 2012 at 3:24 PM, Philip Webb <purslow@ca.inter.net> wrote:
>> >> Regulars will remember the threads re the machine I built recently.
>> >> I thought they mb interested in the start-up time now all is working :
>> >> Gigabyte BIOS  10 s , Linux Lilo prompt - login prompt  8 s ,
>> >> 'startx' - GUI ready  4 s : total  22 s  + entering userid+password ;
>> >> I start the I/net connection (Dhcpcd) manually from the GUI ( 15 s ).
>> >> I assume most of the speed is attributable to the SSD,
>> >> perhaps a bit to the 1600 MHz memory; of course, Gentoo shares the
>> >> honors;
>> >> my desktop manager is Fluxbox & I start apps on desktops manually.
>> >
>> > Toshiba Portégé Z830, with an iCore 5 at 1.60GHz, 6 GB of memory, and
>> > a tiny 128 GB SSD. It takes 12 seconds from GRUB to GDM, and from the
>> > time I enter my password and my GNOME 3 desktop is ready it takes
>> > another 6 seconds, so 18 seconds in total (plus how much it takes for
>> > me to click in my user and enter my password).
>> >
>> > Like you, I attribute most of the speed gain to the SSD. The rest is
>> > systemd.
>> Damn, is GNOME fat. I booted to text console (disabled GDM), and I
>> also disabled plymouth. From GRUB2 to login prompt it takes less than
>> 6 seconds, so the really slow part is starting GDM and then switching
>> to GNOME 3. The BIOS is pretty fast, it takes 4 seconds from power on
>> to the GRUB2 menu.
>>
>> The fast part (GRUB2->login prompt) is because of systemd.
>
> I doubt that,

Install systemd and do the test; I got the numbers to prove it.
systemd is consistently faster than OpenRC (which doesn't even
properly  support parallel starting of services), sometimes several
times faster.

Luca Barbato mentioned about a way to make OpenRC use busybox in
reentrant mode; the difference in speed in that case should be less.
However, the fact is that OpenRC doesn't support parallel start of
services; it said so in its own documentation:

https://bugs.gentoo.org/show_bug.cgi?id=391945#c10

"rc_parallel has never officially been declared a stable feature (see
the comments in rc.conf regarding this)."

So no matter how fast the scripts could execute (which anyway will be
slower than small highly optimized C programs), the lack of proper
parallelization will make OpenRC slower than systemd.

So doubt as much as you want. It doesn't change the fact that (in this
particular issue), you are wrong.

Regards.
-- 
Canek Peláez Valdés
Posgrado en Ciencia e Ingeniería de la Computación
Universidad Nacional Autónoma de México