Speeding up boot process

[Jean-Eric Cuendet]

Hi,
What do you think guys if someone did the following to speed the boot
time in RHL?
- Boot kernel (uncompressible)
- Run the init scripts S00 -> S50
- Start X + XDM + UserLogin
- Continue init scripts S51->S99 in background

It's necessary to start keyboard, mouse, network before letting the user
login.
But it's not necessary to start LDAP, CUPS, MySQL, etc..
It would speed up the boot a lot and is easy to implement.
Do you see problems?
-jec

-- 
Jean-Eric Cuendet
Riskpro Technologies SA
Av du 14 avril 1b, 1020 Renens Switzerland
Principal: [phone]  Fax: [phone]
Direct: [phone]
E-mail: jean-eric.cuendet(a)rptec.ch
http://www.rptec.ch
--------------------------------------------------------

[Bill Nottingham]

John Hearns (John.Hearns(a)micromuse.com) said: 
>  Have a look at the Serel Fastboot project. 
It's been tested. It didn't appreciably help.

Bill

[Mike Chambers]

On Thu, 2003-07-24 at 05:00, John Hearns wrote:

>  Have a look at the Serel Fastboot project.
>  This does a dependency analysis of booting the first time round,
>  then after that sets off the init scripts in a fashion like you describe.
>  
>   http://www.fastboot.org/ 
This looks like it hasn't been developed since last year for Red Hat 7.x
systems.  Or at least was only running for 7.x systems.  I even tried to
get it to work, but during the patch step, it failed.  I'm sure due to
the rc file being different in severn then 7.x's.

-- 
Mike Chambers
Madisonville, KY

"Everyone is entitled to be stupid, but some abuse the privilege."

[Jean-Eric Cuendet]

On Thu, 2003-07-24 at 17:12, Bill Nottingham wrote:
>  John Hearns (John.Hearns(a)micromuse.com) said: 
>  > Have a look at the Serel Fastboot project.
>  
>  It's been tested. It didn't appreciably help. 
OK, but is a speedup of boot process planned?
How? I'm sure a simple solution could be implemented easily.
But it must be an agreed upon one and a one that would have interest from packagers...
Do you have any preferred way to do that?
Or is it just not planned/not useful in your opinion?
-jec

[Simon Perreault]

On July 24, 2003 12:29, Bill Nottingham wrote:
>  It's just all the ways
>  so far tried to speed up the *entire* bootup process (not
>  just get a faster X up) haven't really been successful at
>  all. 
Why isn't the onus on X getting up faster instead? That's what matters.

-- 
Simon Perreault <nomis80(a)nomis80.org&gt;
http://nomis80.org

[Féliciano Matias]

Le jeu 24/07/2003 à 19:09, Simon Perreault a écrit :
>  On July 24, 2003 12:29, Bill Nottingham wrote:
>  > It's just all the ways
>  > so far tried to speed up the *entire* bootup process (not
>  > just get a faster X up) haven't really been successful at
>  > all.
>  
>  Why isn't the onus on X getting up faster instead? That's what matters. 
I don't no if i am the only one but i really don't care about that.
60 seconds or 30 that doesn't matters.
However, i think it's more important to have a suspend mode.

-- 
Féliciano Matias <feliciano.matias(a)free.fr&gt;

[Simon Perreault]

On July 24, 2003 15:40, Mark Mielke wrote:
>  On Thu, Jul 24, 2003 at 08:49:16AM -0400, Simon Perreault wrote:
>  > Jean-Eric Cuendet wrote:
>  > >What do you think guys if someone did the following to speed the boot
>  > >time in RHL?
>  > >- Boot kernel (uncompressible)
>  > >- Run the init scripts S00 -> S50
>  > >- Start X + XDM + UserLogin
>  > >- Continue init scripts S51->S99 in background
>  >
>  > I think this should have been done a long time ago.
>
>  Is that really so easy to do, though?
>
>  X is normally managed by init, wheras the init scripts are
>  managed by /etc/rc.d/rc.sysinit. 
When rc reaches S51, it forks a children which continues the job, detaches it 
and returns. init continues its job and starts X, while rc is still starting 
daemons.

-- 
Simon Perreault <nomis80(a)nomis80.org&gt;
http://nomis80.org

[Mark Mielke]

On Thu, Jul 24, 2003 at 08:49:16AM -0400, Simon Perreault wrote:
>  Jean-Eric Cuendet wrote:
>  >What do you think guys if someone did the following to speed the boot
>  >time in RHL?
>  >- Boot kernel (uncompressible)
>  >- Run the init scripts S00 -> S50
>  >- Start X + XDM + UserLogin
>  >- Continue init scripts S51->S99 in background
>  I think this should have been done a long time ago. 
Is that really so easy to do, though?

X is normally managed by init, wheras the init scripts are
managed by /etc/rc.d/rc.sysinit.

I think you are suggesting that the current /etc/inittab \ /etc/rc.d
scheme be re-invented. Perhaps it has merit.

It isn't a small change though, at least from a maintenance / portability
perspective...

mark

-- 
[e-mail]/[e-mail]/markm(a)nortelnetworks.com __________________________
.  .  _  ._  . .   .__    .  . ._. .__ .   . . .__  | Neighbourhood Coder
|\/| |_| |_| |/    |_     |\/|  |  |_  |   |/  |_   | 
|  | | | | \ | \   |__ .  |  | .|. |__ |__ | \ |__  | Ottawa, Ontario, Canada

  One ring to rule them all, one ring to find them, one ring to bring them all
                       and in the darkness bind them...

                           http://mark.mielke.cc/

[Jean-Eric Cuendet]

On Thu, 2003-07-24 at 18:29, Bill Nottingham wrote:
>  Jean-Eric Cuendet (jec(a)rptec.ch) said: 
>  > On Thu, 2003-07-24 at 17:12, Bill Nottingham wrote:
>  > > John Hearns (John.Hearns(a)micromuse.com) said: 
>  > > > Have a look at the Serel Fastboot project.
>  > > 
>  > > It's been tested. It didn't appreciably help.
>  > 
>  > OK, but is a speedup of boot process planned?
>  > How? I'm sure a simple solution could be implemented easily.
>  
>  Have you tried? If so, we'd love to hear simple solutions. :)
>  
>  It's not that we're not interested. It's just all the ways
>  so far tried to speed up the *entire* bootup process (not
>  just get a faster X up) haven't really been successful at
>  all. 
IMO, the important thing is not to get the system up quickly, it's to
get X up and running quickly.
It's to be able to login quickly, not that the last service started is
available.
Take WinXP, they boot very quickly to allow the user to lgin quickly,
not that SQL Server is available quickly. And THAT makes the difference.
-jec

[Bill Nottingham]

Jean-Eric Cuendet (jec(a)rptec.ch) said: 
>  On Thu, 2003-07-24 at 17:12, Bill Nottingham wrote:
>  > John Hearns (John.Hearns(a)micromuse.com) said: 
>  > > Have a look at the Serel Fastboot project.
>  > 
>  > It's been tested. It didn't appreciably help.
>  
>  OK, but is a speedup of boot process planned?
>  How? I'm sure a simple solution could be implemented easily. 
Have you tried? If so, we'd love to hear simple solutions. :)

It's not that we're not interested. It's just all the ways
so far tried to speed up the *entire* bootup process (not
just get a faster X up) haven't really been successful at
all.

Bill

[Owen Taylor]

On Thu, 2003-07-24 at 11:47, Jean-Eric Cuendet wrote:
>  On Thu, 2003-07-24 at 17:12, Bill Nottingham wrote:
>  > John Hearns (John.Hearns(a)micromuse.com) said: 
>  > > Have a look at the Serel Fastboot project.
>  > 
>  > It's been tested. It didn't appreciably help.
>  
>  OK, but is a speedup of boot process planned?
>  How? I'm sure a simple solution could be implemented easily.
>  But it must be an agreed upon one and a one that would have interest from packagers...
>  Do you have any preferred way to do that?
>  Or is it just not planned/not useful in your opinion? 
After having tried various things, and timed them,
we tend to be a lot less confident that a simple solution
could be implemented easily.

The vast majority of startup time seems to be reading stuff
of the disk; parallelizing init scripts doesn't
help with that. Starting up GDM earlier may work, but
with the downsides that:

 - You have to figure out what services are safe to
   start after login.
 - You may have a unresponsive system during login.

https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=startuptime

is a recently created tracker bug for this issue.

Regards,
					Owen

[Simon Perreault]

Jean-Eric Cuendet wrote:

> Hi,
> What do you think guys if someone did the following to speed the boot
> time in RHL?
> - Boot kernel (uncompressible)
> - Run the init scripts S00 -> S50
> - Start X + XDM + UserLogin
> - Continue init scripts S51->S99 in background
>  
I think this should have been done a long time ago.

[John Hearns]

>  Hi,
>  What do you think guys if someone did the following to speed the boot
>  time in RHL?
>  - Boot kernel (uncompressible)
>  - Run the init scripts S00 -> S50
>  - Start X + XDM + UserLogin
>  - Continue init scripts S51->S99 in background
>  
Have a look at the Serel Fastboot project.
This does a dependency analysis of booting the first time round,
then after that sets off the init scripts in a fashion like you describe.

 http://www.fastboot.org/

I played with this last year, and it seemed fine on my laptop.
Sadly it didn't play that well with the standard CERN environment, which
uses AFS.
Not sure what the exact problem was, but I didn't pursue it.