{OpenSuSE, Fedora, Mandriva, Ubuntu}- how light can they be?

I am using Linux since 2002. I started with Mandrake, but quickly moved to Vector and Slackware (stayed on Slackware till 2008th). I can say that I prefer fast, well optimized and efficient distros, but because of certain reasons I had to move to Fedora. Now my laptop (my main and only machine) runs Fedora 10 and my sister’s laptop runs OpenSuSE 11.1.

Two days ago, I had some free time so I read a little about Sidux and decided to test it on one partition of my sister’s laptop. After full upgrade I got kernel 2.6.30 and KDE 4.2.4. The thing that amazed me (in the same time thing that made me write this topic) is how fast it was. My laptop is quite decent machine (AMD Turion x64 2.0 GHz with 4 gigs of RAM), it runs Fedora 10 i386 with KDE4 with minimum of services and minimal possible install (for my needs, of course) and it felt so sluggish and bloated compared with Sidux on less powerful machine (basically the same amount of software and services). As far as I know, Sidux is using Debian Sid branch and Debian has no special optimizations, generic ones like mainstream distros.

So the question is - how much is possible to make OpenSuSE (or any other distro from the name of the topic) light, fast and efficient? My sister’s install is basically custom (installed server system and then manually using zypper I installed the rest), but still slugish comapring to Sidux default install (KDE-lite tough) + additional software.

Thank you for your answers!

> So the question is - how much is possible to make OpenSuSE (or any
> other distro from the name of the topic) light, fast and efficient?

that is kinda like asking:

Q: Of all the automobiles manufactured today, how much is possible to
make them light, fast and efficient?

A: That depends on how much time and effort at rebuilding you intend
to invest. As well as the level of your particular mechanical skills
and tools set.

so, back to your question and my A: All can be slimmed down to be
lighter, faster and more efficient.

another way to attack this problem would be to just pick an already
light, fast and efficient distro and do NOT load it down trying to
make it “look better” and have more eye-candy, bells and whistles and
etc than the Redmond Pig. (which, imho, seems to be the focus of the
current openSUSE/KDE4 development clique…)

ymmv,

brassy

I perfectly understand what are you talking about, but I meant how light, fast and efficient can they be with default packages. Of course, it is always possible to rebuild something with additional optimizations, but that is not quite smart and logical, since it is much easier to simply switch to another distro.

I.e. I’ve read few months ago that many people used alternate Ubuntu install CD and they built it manually (by using APT) installing just necessary stuff and results were quite amazing. That is why I’m interested if there are similar experiences with OpenSuSE.

Also, I forgot to mention RAM usage:

  1. sidux install with KDE 4.2.4 and ktorrent running 3 torrents, took only (believe it or not) 130 MB (I used free -m command).
  2. for comparision, Fedora with KDE 4 and without additional software running is starting basically with 250 MB.

kukibl wrote:
> I meant how light, fast and efficient can they be with default packages.

define “default packages”!

to begin with there are <on http://software.opensuse.org/> THREE
choices (32 bit, 64 bit and Power PC) then you can choose TWO disks
(CD and DVD)…

that give SIX “default packages”…

AND, if you select CD above, THEN you can choose between Gnome or
KDE4…giving, 3x2 more options…

BUT, if you select the DVD as your “default”, THEN during install you
can select from Gnome, KDE3, KDE5, XFCE, Minimal X Window, Minimal
Server Selection, giving 3x6 options…

THEN, you can later in the install process you can add or delete the
pigs you wish…

the answer to your question depends on what you consider “default”.

generally, they are faster, lighter and more efficient than sliced
bread…next question!

but, for a more definitive answer i’d suggest you install all the
possible defaults, measure the weight, speed and efficiency or each
and stuff the data into a OpenOffice.org Calc chart, complete with
colored graphs…


brassy

Default packages - packages originally made by OpenSuSE developers! I already wrote that it is always possible to pull down source packages and rebuild them with certain optimizations, put that is not topic of this post. All possible choices you wrote are actually different combinations of same groups of packages, so there are no differences between them because you can always “play” with zypper a little bit.

That is why I started this topic, to avoid that kind of experiment. It is time consuming and probably at the end of the day not that much necessary. I want to hear what other people experienced, I am sure that not everybody is using default choices (DVD, live CDs install) and add only software they need. I believe that there are lots of people who play with default selections, remove unnecessery packages and services, maybe even modify configuration files in order to get less bloated system.

Tweaking can only do so much.

Except for beagle, I don’t find that turning off daemons does much, unless you are desperately short of memory. Their footprints are in the order of tens of MB. Firefox’s footprint easily swamps that.

Removing packages does nothing except free up some disk space, which is cheap these days. If you never use OO, it makes no difference removing its packages.

Fast graphics cards, fast disks and controller, fast CPU, and most important, sufficient memory help a lot.

Compilation optimisation a la Gentoo may get you some increase in speed but I’m not prepared to have my machine grind away for hours just to get it. It may be worth it for some CPU intensive programs like encoders and media players, but those already use tricks to select the best code for the CPU.

I’m not willing to use lighter but less capable versions of software just to be able to run faster, if it means I have to spend more time coping with lack of features I want. With hardware so cheap these days, over the lifetime of the computer the time wasted outvalues the saving in using a slower machine.

I couldn’t agree more with everything you wrote.:wink: …but - could somenone explain me why such difference between sidux and Fedora with basically same install (no additional services or daemons)? 100 MB is not small difference.

Right now I’m running my sister’s laptop (OpenSuSE 11.1) with minimal KDE3 and Firefox. It still looks slower compared to sidux with KDE4, but this impressed me:


$free -m
             total       used       free     shared    buffers     cached
Mem:          1876       1777         98          0         90       1551
-/+ buffers/cache:        135       1740
Swap:          251          0        250

as you can see, most of your memory is cached which is a good thing as free memory is useless and also it is faster when apps are cached in memory compared to always loading them from disk

further, KDE4 has a smaller memory footprint than KDE3 so a comparison here is not valid. Maybe Sidux compiles everything with -finline-functions which can make software run ~10% faster and which SUSE 11.2 will also use for all its packages. Presonally, on my machines, SUSE is very fast for me

I am not checking cached memory, just -/+ buffers/cache. I am quite impressed that even with Firefox running it is still beneath 150 megs.

Yes, recent versions of OpenSuSE (primaryly 11 and 11.1) have very good performance for me too. I am not that sure that sidux compiles many packages, maybe just some core stuff (like kernels), all other packages are part of Debian Sid branch and I already wrote that I doubt that Debian has any special optimizations. But you never know…

Anyway, my comparision was between Fedora and sidux, both running KDE4.

You just quoted some numbers for sidux and Fedora without saying how you measured them. You should be looking at the line:

-/+ buffers/cache

and without any apps running in the GUI. Also on the exact same hardware, because the footprint of the X server depends on what drivers are used. Also some of the video memory that’s mapped in is also counted in the X server’s footprint. So unless you are careful in what you are measuring, that 100MB may be significant, or not.

Yes, as I wrote in my previous post I am just cheking that line : -/+ buffers/cache.

sidux is running on less powerful machine with Radeon graphics (radeonhd driver) with ktorrent (running 3 torrent) and it was arround 130 MB. Fedora is running on more powerful machine with Radeon graphics (fglrx driver) with only KDE4 (no additional apps) and it was arround 250 MB. Both system are x86. Xorg server version is 1.6 in sidux and I believe 1.5 in Fedora.

@microchip8

Very nice repository. Many useful stuff I found there.

You mean -fomit-frame-pointer? Note that it is enabled already for x86-64, the 11.2 change only applies to i586.

About the question… you are the one with a Sidux installation. You are the one better fitted to answer your question.

A different graphics card could mean that it’s running with a different driver, and in fact you are. If you are going to be doing benchmarking, you need a level playing field. Same hardware, as far as possible, same versions of programs, except the OS of course.

Or just sit back, enjoy it and be happy.

Hey… level4 is waiting for you!

This is what i know (only the surface) so… anyone out there with a deeper knowledge : fell free to provide more informations.

All Linux distributions are ‘one size fits all’ approach. By doing this it doesn’t provide the ‘optimized system’. (you may never use ‘the plutonium mouse’ but it will remain in the subsytem memory, by removing it you not only free that memory but you also speed up boot times.

Create custom run-level (rl4 is there exactly for that)

all under root privileges

clear rl4 (if there is) scripts with: rm /etc/init.d/rc4.d/*

create a clone of rl5 to rl4 with: cp -P /etc/init.d/rc5.d/* /etc/init.d/rc4.d/

edit /etc/inittab file. Gnome: gedit /etc/inittab KDE: kate /etc/inittab

change the 5 to a 4

scroll down to the line: #l4:4:wait:/etc/init.d/rc 4

remove the hash

save and quit editor

reboot

if its works, (if its not working restore rl5) now is the time to prune run-level services (use the YaST tool)

to restore rl5:

reboot, GRUB menu take Failsafe

at loging use the root option

type : pico /etc/inittab to open the : /etc/inittab file

go down the line that reads: id:4:initdefault: (change 4 to 5)

save (Ctrl+O(letter))

Ctrl+X to exit pico editor

type reboot

quick review:
Understand bootup
reduce boot menu delay
optimize run-level services
create custom run level
prune run-level services
if its not working, restore run-level 5

try it and enjoy a leaner system.

mmmm, the benefits of having a system that can own the full package. :stuck_out_tongue: Seriously, though, run level 4 is great. I use it on my laptop and it’s swift. I give it two thumbs up.

I understand. Just for curiosity, as soon as I finish downloading Fedora 11, I’ll replace sidux install with it and check for differences. Same machine, same drivers, same apps, different OS.:slight_smile:

@Roger_Marquis

Very interesting idea, definitely I’ll give it a try as soon as I find enough free time. Thank you!