Speed

Questions about the project and the distribution - obviously no support questions here please

Speed

Postby AlsaPhil on Mon Jan 15, 2007 12:26 pm

The boot times represent the number of seconds each distribution took to get from LILO or GRUB boot prompt to the KDE or GNOME login prompt in their default installations.

Boot Times
Distribution Seconds
Mandriva Linux 2007 40
PCLinuxOS 0.93a 46
Debian GNU/Linux 4.0 47
Pardus Linux 2007 48
Ubuntu 6.10 49
SimplyMEPIS 6.0-4 (Beta 1) 51
openSUSE 10.2 61
Fedora Core 6 72
Xandros Desktop 4.1 89


(...)


Seen today on distrowatch's weekly news. The boot time is between 40" and 89" on these examples.
My experience with Mint gives 78"/80" for the same way GRUB > GNOME login (on 3 separated laptops).

It is not a big problem I just would like to know where the difference could be :oops:

At second, is it possible to do it better?

For information, I installed an Ubuntu Herd CD2 this morning just to see what it is, boot time 40" on one of my 3 laptops.
Image
User avatar
AlsaPhil
Level 4
Level 4
 
Posts: 397
Joined: Thu Nov 16, 2006 9:32 am
Location: Tunis, Tunisia

Linux Mint is funded by ads and donations.
 

Re: Speed

Postby scorp123 on Mon Jan 15, 2007 12:40 pm

AlsaPhil wrote:It is not a big problem I just would like to know where the difference could be
Services. The more services you run the longer the boot will take. The strategy would be to remove unneeded services (but remove wisely or else you'll criple your machine!) until you're satisfied with the speed. Other factors (e.g. hardware, network speed, congestion, disk partitioning, disk order, disk fragmentation, etc.) also play an important role.
User avatar
scorp123
Level 8
Level 8
 
Posts: 2287
Joined: Sat Dec 02, 2006 4:19 pm
Location: Switzerland

Postby AlsaPhil on Mon Jan 15, 2007 12:47 pm

The more services you run the longer the boot will take


Thanks for answer, scorp :)
Does it mean that Mandriva or open Suse (for instance) run less services?
Does it mean Tomboy and Beagle are "using" 30"?
Once I tried removing Beagle... Result was same
:oops:
Image
User avatar
AlsaPhil
Level 4
Level 4
 
Posts: 397
Joined: Thu Nov 16, 2006 9:32 am
Location: Tunis, Tunisia

Postby scorp123 on Mon Jan 15, 2007 1:00 pm

AlsaPhil wrote:Does it mean that Mandriva or open Suse (for instance) run less services?
The opposite is the case, they are running services like mad. But they are cheating a little, e.g. I know that SUSE is using a program called preload ... Everytime you shutdown a SUSE system it will calculate stuff about some of the binaries that will get started next time, so that said binaries will load fast like hell the next time they are loaded into memory, e.g. when the next reboot happens. Another difference is that e.g. SUSE binaries are always compiled for i686 instruction code (= the latest CPU generation) whereas Debian still uses old i386 (= Intel 80386 CPU) instruction sets to maintain max. compatibility with all the Intel-compatible CPU's out there.

The fastest ever boot I have seen was a Gentoo installation compiled from the ground up with NO services running whatsoever and booting straight into a KDE session ... That thing booted in under 10 seconds. But the machine was basically useless. It wasn't running any services, not even loading sound drivers during boot. The point was just to show that it could be done. :D
Last edited by scorp123 on Mon Jan 15, 2007 2:31 pm, edited 2 times in total.
User avatar
scorp123
Level 8
Level 8
 
Posts: 2287
Joined: Sat Dec 02, 2006 4:19 pm
Location: Switzerland

Postby AlsaPhil on Mon Jan 15, 2007 1:14 pm

I understand, thanks.
For your last example, I think we can do it still more faster :roll: without Gentoo :lol:
Image
User avatar
AlsaPhil
Level 4
Level 4
 
Posts: 397
Joined: Thu Nov 16, 2006 9:32 am
Location: Tunis, Tunisia

Postby scorp123 on Mon Jan 15, 2007 2:34 pm

AlsaPhil wrote:For your last example, I think we can do it still more faster :roll: without Gentoo :lol:
For BeOS it was absolutely normal to boot in less than 10 seconds into the desktop, with all filesystem, graphics and sound drivers loaded! And we're talking Pentium II class machines here (around 400 or so MHz).
User avatar
scorp123
Level 8
Level 8
 
Posts: 2287
Joined: Sat Dec 02, 2006 4:19 pm
Location: Switzerland

Postby 900i on Mon Jan 15, 2007 4:35 pm

Last edited by 900i on Tue Jan 16, 2007 6:09 am, edited 1 time in total.
Intel i7 3770 LMDE Mate 64
User avatar
900i
Level 6
Level 6
 
Posts: 1014
Joined: Sat Nov 18, 2006 9:30 am
Location: Wakefield, UK

Re: Speed

Postby scorp123 on Mon Jan 15, 2007 6:29 pm

AlsaPhil wrote:Xandros Desktop 4.1 89
That one doesn't surprise me one little bit ... Xandros is such a horrible mess! :roll: The Xandros boot strategy is to simply start every damn script that is installed in the system's init folder, whether it's needed or not, wether the device is there or not. For example it tries to start scripts that would initiate some drivers specific to Sony VAIO laptops even though you're not even running this stuff on a Sony. And so it wastes a lot of time by trying to start unneeded scripts. A total mess IMHO.

You can easily verify this by installing Xandros and then by disabling their ugly swastika-like boot splash ... you'll see tons and tons of messages from totally unneeded boot scripts fly by.

They really could have done a better job there :?
User avatar
scorp123
Level 8
Level 8
 
Posts: 2287
Joined: Sat Dec 02, 2006 4:19 pm
Location: Switzerland

Postby clem on Tue Jan 16, 2007 5:33 am

hmm... interesting :)

I suppose, if you don't have an Intel graphic card you might be able to remove the package 915resolution which gets loaded at startup.... that might speed things up.

:roll: :oops: :oops:
User avatar
clem
Level 15
Level 15
 
Posts: 5553
Joined: Wed Nov 15, 2006 8:34 am

Postby AlsaPhil on Tue Jan 16, 2007 6:47 am

... but I have :lol:
Image
User avatar
AlsaPhil
Level 4
Level 4
 
Posts: 397
Joined: Thu Nov 16, 2006 9:32 am
Location: Tunis, Tunisia

Postby Josh on Thu Jan 18, 2007 3:46 pm

scorp123 wrote: The opposite is the case, they are running services like mad. But they are cheating a little, e.g. I know that SUSE is using a program called preload ... Everytime you shutdown a SUSE system it will calculate stuff about some of the binaries that will get started next time, so that said binaries will load fast like hell the next time they are loaded into memory, e.g. when the next reboot happens. Another difference is that e.g. SUSE binaries are always compiled for i686 instruction code (= the latest CPU generation) whereas Debian still uses old i386 (= Intel 80386 CPU) instruction sets to maintain max. compatibility with all the Intel-compatible CPU's out there.


FWIW, Ubuntu has a very similar tool to SuSE's preload. It's called readahead. Both essentially do the same thing in slightly different ways. Here's a basic tutorial on using readahead if you're interested. There is also the sysvinit replacement that edgy uses, which helps to decrease boot time - Upstart. Upstart is a really cool idea and is improving constantly. Read through some of the specs in the wiki to see what I mean.

Also, 99% of all Ubuntu packages (perhaps higher) are compiled for i486. Not that it makes a hoot of difference - the few clock cycles saved on your processor by compiling for i686 wouldn't even be noticable, and definitely wouldn't improve boot times by any serious margin. The fact is that most average end-user applications simply do not code in processor specific optimizations - meaning compilers can only do so much. Those little optimizations that are done are so insignificant compared to I/O wait times, network responses, user inputs, etc. as to be basically meaningless.
User avatar
Josh
Level 1
Level 1
 
Posts: 18
Joined: Sat Nov 18, 2006 10:29 am
Location: Abilene, TX USA

Postby scorp123 on Thu Jan 18, 2007 4:02 pm

Josh wrote: The fact is that most average end-user applications simply do not code in processor specific optimizations - meaning compilers can only do so much. Those little optimizations that are done are so insignificant compared to I/O wait times, network responses, user inputs, etc. as to be basically meaningless.
Absolutely. And also real-time processing as requested by another user in another thread wouldn't help much. The only thing real-time processing does is that it enables you to (fore)tell precisely how long a certain process will need to get started and to shutdown. But it doesn't do any magic and speed up your system. If you still got too many unnecessary services starting during boot then real-time processing won't change much, if anything at all.
User avatar
scorp123
Level 8
Level 8
 
Posts: 2287
Joined: Sat Dec 02, 2006 4:19 pm
Location: Switzerland

Postby Josh on Thu Jan 18, 2007 5:18 pm

Yep, the simplest thing any user can do is slim down the scripts called at boot. Getting rid of kernel modules and compiling a monolithic kernel is another idea I have seen people discuss. But that has it's own set of drawbacks for most people, and IME again can only provide minute gains at the absolute best.

Now, once you're up and running there are lots of little tweaks here and there that can help. I'm more experienced with the kernel itself, but I know of several 'hacks' to speed up one's desktop. And if you're doing video or music mixing/sampling RTS (realtime systems) are an essential.

Sorry, I just get all geeky discussing these things :D
User avatar
Josh
Level 1
Level 1
 
Posts: 18
Joined: Sat Nov 18, 2006 10:29 am
Location: Abilene, TX USA

Postby .ee on Mon Feb 19, 2007 11:54 am

I would like better boot up and hibernation time for Mint. I removed the languages and some services but it is still around 40 sec. Well. I have 4200rpm 1,8" had disk with up to 20MB/s transfer rate, so someone may have better boot up times without doing anything. Still, my optimized Windows XP boots and awakes 30% faster! I don't see reason why Mint sould not be able match Windows bootup times.

How I can benchmark my HDD transfer rates under linux, to be sure it performs equally well under linux? I mean the utility like HDTune in Windows or something run from the command line at least.

I am also concerned that my hard disk sound slightly noisier under linux.
This is the reason why I want to check the performance. Maybe it does not work optimally.

Any ideas?
.ee
Level 2
Level 2
 
Posts: 55
Joined: Mon Feb 05, 2007 9:52 pm

Linux Mint is funded by ads and donations.
 

Return to Non-technical Questions

Who is online

Users browsing this forum: No registered users and 3 guests