LinuxQuestions.org - General question about Debian Performance

- Debian (https://www.linuxquestions.org/questions/debian-26/)

- - General question about Debian Performance (https://www.linuxquestions.org/questions/debian-26/general-question-about-debian-performance-642445/)

General question about Debian Performance

Hello Everyone,

I just have a couple of questions. I have a number of distros dual booting on my desktop, one of them being Debian Etch. The others are PCLinuxOS, Fedora 9, and Ubuntu Hardy (as well as windows XP,) For a number of weeks I haven't booted into Debian as I was doing some projects in Ubuntu and Fedora. Today I booted into Debian, and there were 51 updates. One of them was the same kernel that I have installed (2.6.18.6-i486.)

I was just wondering why there would be an update to the same kernel. The update of all the packages went fine, but I was informed that a number modules would need to be rebuilt. Was there a patch to this kernel that was necessary for all systems?

Also, since I have not been using Debian for a while, I was wondering what it is that makes Debian so much faster and smoother to run on the same hardware? I have virtually the same software installed on all of the distros mentioned above, including Debian, but the computer runs cooler, much faster and generally smoother all the way around. I remember this also being true the first time I tried Debian about three years ago. I use all of these distros for different purposes but none of them run as quickly as Deb. As a matter of fact, Ubuntu stutters sometimes on this computer, when Debian just flies. Does anyone know an easy answer as to why the Debian build is so fast?

Just interested for future reference in projects I am considering.

Thanks in advance,

Bob

Quote:

Originally Posted by BobNutfield (Post 3154390)

The update to the same kernel that you saw was because of some security patches that were released by Debian Security team. Generally once the testing becomes stable in Debian, there are no more updates pushed in stable except some serious security issues. The update you saw was a security issue in the kernel. It is same to install that. Make sure you boot as soon as the installation finishes.

I have noticed this also. I have assumed that it was because Debian is "no frills". But Arch and Slackware are also pitched as no-frills distros and they are slower.

I'd like to see generic answers to this---ie what are the major factors that influence speed of a Linux installation?

Quote:

Make sure you boot as soon as the installation finishes.

Kushakoolwal: Thank you. I did reboot immediately after update and all rebooted fine. I assumed in was a security update but that was not described in the update.

pixellany: I actually did a boot time test, measured from selected at Grub to login:

Fedora: 33 seconds
Ubuntu: 41 seconds
PCLinux 28 seconds
WinXP 34 seconds

Debian 16 seconds

Quote:

Originally Posted by BobNutfield (Post 3154414)

One of the reasons is that other distros like Fedora and Ubuntu are like Vista of Linux world. They have everything that you want and everything that you don't want. Their default installation is heavily bloated. Check by running "ps -ef" and you will see a lot more processes/services running.

Yes, I know that is the case, particularly in Ubuntu (not my favorite distro right now). I believe in the case of Fedora the default SELinux installation really slows things down. I got a reasonably good speed increase when I totally disabled it. When I did my boot up test, I removed the splash screens on all of them (except Windows, of course, because I couldn't) so that I nothing but text scrolling across the screen. PCLinux is pretty quick, but no where near as quick as Debian.

Bob

Just for grins, you CAN disable the Splash screen in Windows XP if you want.
You could also get to see the drivers etc loading in windows, similar to a Linux boot with no splash.

http://windowsxp.mvps.org/nosplash.htm

That would be interesting. How would one do that?

Bob

One reason that Debian does so well is that it has a kernel that is specifically built for each architecture on which it runs. Another reason is that if you don't add a bunch of junk to a default Debian instalation, the base software is relatively lean and it does not start up too many services.

With that in mind, I did formerly have a Debian system set up, based on Debian Etch, and I added a lot of software to it. While it still ran relatively well, a more modest sized Debian derivative distribution that does not have a lot of extras would run better. Examples: SimplyMEPIS runs very well, one of the best moderate sized simple desktop systems, based on Debian. DreamLinux has also run very well in the past.

All that said, a simple, no frills Debian system is one of the best systems you can have. Debian derived systems can be very good too. My two favorites in this category are sidux for a Debian Sid, cutting edge system that actually works, and SimplyMEPIS for a basic, stable system that you can count on to work. One of the even smaller SimplyMEPIS derivatives, which grabs some of the smallest and best from MEPIS, and hence from Debian itself is AntiX. You will find it to be a fast and faithful Debian implementation as well. If you have old hardware, look into AntiX. You will find that Debian feel to it.

Quote:

Originally Posted by BobNutfield (Post 3154414)

Slackware 12.1: 52 seconds (to the kdm login prompt)
I have more research to do.....;) (For example, my test was on a laptop)

Is comparing bootup times really worth anything? I mean as a true measurement of speed?

Quote:

Is comparing bootup times really worth anything? I mean as a true measurement of speed?

Since the various OSes are all on the same hardware, some conclusions can be drawn, but not conclusions that would translate actual processing efficiency from one Distro to the next. I like to play around with Super_Pi. Even then, such comparisons are of questionable value. I dual boot Debian 32-bit and 64-bit on this hardware:

Code:

debian32:~$ infobash

CPU[Dual Intel Core2 Duo E6550 @ clocked at 2333.325 Mhz]  Kernel[Linux 2.6.24-1-686 i686]  Up[-2min-]  Mem[-94.9/2026.9MB-]  HDD[-640GB(23%used)-]  Procs[-94-]  Client[Shell]

debian32:~/super_pi$ infobash -v2 0

Host/Kernel/OS  "debian32" running Linux 2.6.24-1-686 i686 [ Debian GNU/Linux lenny/sid ]

CPU Info        (1) Intel Core2 Duo E6550 @ clocked at [ 2333.325 MHz ]

                (2) Intel Core2 Duo E6550 @ clocked at [ 2333.325 MHz ]

Videocard      nVidia G70 [GeForce 7600 GT]  X.Org 1.4.0.90  [ 1280x1024@50.0hz ]

                nVidia G70 [GeForce 7600 GT]

Network cards  2x nVidia MCP55 Ethernet, at ports: ac00 a800 

Processes 93 | Uptime 2min | Memory 98.0/2026.9MB | HDD ATA WDC WD3200AAKS-0,ATA WDC WD3200AAKS-2 Size 640GB (23%used) | GLX Renderer GeForce 7600 GT/PCI/SSE2 | GLX Version 2.1.2 NVIDIA 169.12 | Client Shell | Infobash v3.01

debian32:~$

Super_Pi figures "pi" out to a million or so decimal places.

Code:

debian32:~/super_pi$ sh super_pi 20

 Version 2.0 of the super_pi for Linux OS

 Fortran source program was translated into C program with version 19981204 of

 f2c, then generated C source program was optimized manually.

 pgcc 3.2-3 with compile option of "-fast -tp px -Mbuiltin -Minline=size:1000 -Mnoframe -Mnobounds -Mcache_align -Mdalign -Mnoreentrant" was used for the

 compilation.

 ------ Started super_pi run : Sat May 17 11:29:10 MDT 2008

 Start of PI calculation up to 1048576 decimal digits

 End of initialization. Time=      0.244 Sec.

 I= 1 L=      0        Time=      0.716 Sec.

 I= 2 L=      0        Time=      0.816 Sec.

 I= 3 L=      1        Time=      0.812 Sec.

 I= 4 L=      2        Time=      0.816 Sec.

 I= 5 L=      5        Time=      0.816 Sec.

 I= 6 L=      10        Time=      0.820 Sec.

 I= 7 L=      21        Time=      0.820 Sec.

 I= 8 L=      43        Time=      0.816 Sec.

 I= 9 L=      87        Time=      0.812 Sec.

 I=10 L=    174        Time=      0.816 Sec.

 I=11 L=    349        Time=      0.816 Sec.

 I=12 L=    698        Time=      0.816 Sec.

 I=13 L=    1396        Time=      0.820 Sec.

 I=14 L=    2794        Time=      0.812 Sec.

 I=15 L=    5588        Time=      0.812 Sec.

 I=16 L=  11176        Time=      0.804 Sec.

 I=17 L=  22353        Time=      0.792 Sec.

 I=18 L=  44707        Time=      0.760 Sec.

 I=19 L=  89415        Time=      0.712 Sec.

 End of main loop

 End of calculation.    Time=      16.073 Sec.

 End of data output.    Time=      0.108 Sec.

 Total calculation(I/O) time=      16.181(      0.576) Sec.

 ------ Ended super_pi run : Sat May 17 11:29:26 MDT 2008

Again, on a fresh boot, with the 64-bit running a newer kernel...

Code:

debian64:~$ infobash -v2 0

Host/Kernel/OS  "debian64" running Linux 2.6.25-1-amd64 x86_64 [ Debian GNU/Linux lenny/sid ]

CPU Info        (1) Intel Core2 Duo E6550 @ clocked at [ 2333.327 MHz ]

                (2) Intel Core2 Duo E6550 @ clocked at [ 2333.327 MHz ]

Videocard      nVidia G70 [GeForce 7600 GT]  X.Org 1.4.0.90  [ 1280x1024@60.0hz ]

                nVidia G70 [GeForce 7600 GT]

Network cards  2x nVidia MCP55 Ethernet, at ports: ac00 a800 

Processes 93 | Uptime 1min | Memory 109.0/2014.3MB | HDD ATA WDC WD3200AAKS-0,ATA WDC WD3200AAKS-2 Size 640GB (23%used) | Client Shell | Infobash v3.01

debian64:~$



debian64:~/super_pi$ sh super_pi 20

 Version 2.0 of the super_pi for Linux OS

 Fortran source program was translated into C program with version 19981204 of

 f2c, then generated C source program was optimized manually.

 pgcc 3.2-3 with compile option of "-fast -tp px -Mbuiltin -Minline=size:1000 -Mnoframe -Mnobounds -Mcache_align -Mdalign -Mnoreentrant" was used for the

 compilation.

 ------ Started super_pi run : Sat May 17 11:33:36 MDT 2008

 Start of PI calculation up to 1048576 decimal digits

 End of initialization. Time=      0.252 Sec.

 I= 1 L=      0        Time=      0.716 Sec.

 I= 2 L=      0        Time=      0.812 Sec.

 I= 3 L=      1        Time=      0.816 Sec.

 I= 4 L=      2        Time=      0.812 Sec.

 I= 5 L=      5        Time=      0.812 Sec.

 I= 6 L=      10        Time=      0.812 Sec.

 I= 7 L=      21        Time=      0.824 Sec.

 I= 8 L=      43        Time=      0.816 Sec.

 I= 9 L=      87        Time=      0.812 Sec.

 I=10 L=    174        Time=      0.816 Sec.

 I=11 L=    349        Time=      0.808 Sec.

 I=12 L=    698        Time=      0.812 Sec.

 I=13 L=    1396        Time=      0.832 Sec.

 I=14 L=    2794        Time=      0.808 Sec.

 I=15 L=    5588        Time=      0.812 Sec.

 I=16 L=  11176        Time=      0.804 Sec.

 I=17 L=  22353        Time=      0.788 Sec.

 I=18 L=  44707        Time=      0.764 Sec.

 I=19 L=  89415        Time=      0.716 Sec.

 End of main loop

 End of calculation.    Time=      16.065 Sec.

 End of data output.    Time=      0.104 Sec.

 Total calculation(I/O) time=      16.169(      0.668) Sec.

 ------ Ended super_pi run : Sat May 17 11:33:52 MDT 2008

...the 64-bit is considerably slower than 32-bit. I have no idea why. Maybe the Super_Pi program is somehow optimized for 32-bit.

At any rate attempting head-to-head comparisons of various systems has generally given me a headache, and unsatisfying conclusions.

Quote:

Originally Posted by unSpawn (Post 3156300)

Is comparing bootup times really worth anything? I mean as a true measurement of speed?

I know that it seems like hair-splitting, but--when I sit down in the cafeteria for lunch (one of the few times I can check e-mail during the day)--the difference between 20-second boot and 1 minute is psychologically significant. And--if I'm showing someone how to do something and I have to re-boot in the middle of it, then 1 minute startup is an eternity.

I also instinctively believe that startup time and typical application loading time are correlated. No proof, however.

Arch Linux 2008.04 RC, with kdemod (kdebase only):
23 seconds
20 seconds when the network daemon is disabled (it was looking for wireless which is not yet configured)

Quote:

I wholly agree with this. There are many, many posters in this forum who are experts in both Linux and computers in general. I am not one of them. I do not make my living in anything IT related and have never had any formal training in the field. I simply aquired an interest in Linux in 1998 with one of the old Red Hat boxed retail sets. When I installed it on my old AMD K6 350Mhz (which I temporarily removed Win98 from), the only thing I could make any sense of is that it was amazing to me how fast this thing was booting to the "desktop" (such as it was at that time.) I didn't do any kind of measuring of course, but it was obvious that whatever this thing was, it was fast! I have and use currently 6 distros on five different computers (1 desktop and four laptops) and since computers and Linux are simply my passion and hobby, I look at things and study certain parts of the experience that others who use their systems for much more serious endeavors would never bother with. Debian is on my desktop which is a single core AMD64 (5 years old) with 1.5GB memory and an old FX5500 256mb graphics card. A couple of the laptops are much higher spec and I only have Hardy and Fedora 9 installed on them. But, I do have very similar software installed on EVERY distro, particularly video editing software like Cinelerra. So, as a non-expert hobbyist, I am convinced that there is definitely a correlation between boot times and overall speed of the distro itself in dealing with the same software. Debian seems to scream through these software packages, when the others distros sometime take foreever just to open them. These are things that a non-expert will notice. And, as a previous poster has pointed out, the reason for this probably the absence in Debian of so many of the things that make these other distros "just work" on most hardware.

However, for a non-expert like me, there is a trade off. A case in point is the experience I had getting Linux installed on a new laptop I purchased about a month ago. I tried more than a dozen live distros trying to find one that would recognize and enable the sound and wireless in this machine. Ubuntu was the ONLY one that would recognize all the hardware and get me running right "out of the box". But, it is slow and buggy on this dual core laptop with 2GB memory. There is no doubt in my mind that Debian would have been a much better choice to run on this higher spec laptop, but I know it would have taken me ages to get all the hardware running, if I ever did.

So, I put up with the bloat and occassional freeze because everything just works. The same is true about Fedora 9 on another laptop, though it is much zippier than Ubuntu.

I guess the sum total of what I am saying is that most of us just accept the bloat that Debian doesn't have to get our hardware working with the least amount of fuss.

Bob