Comparing Distribution Boot Times
August 23, 2007 around 6am (openSUSE)I’ve been interested in the way different distributions handle boot time lately, and with openSUSE’s recent and very impressive boot time improvements I thought I’d make a few extra comparative statistics using bootchart.
Rules
To ensure that these results are reasonably fair, I’m posing a few restrictions, most of which are pretty obvious:
- Use the same setup, my laptop, a Vaio VGN-FE11S.
- Local user, IP by DHCP
- Completely default setup, no system updates. As the openSUSE article mentions, updates can really change the speed of boot time since the files you need to access will be fragmented over the hard disk.
- Time recorded will have to be the time until the KDM screen loads, with no auto-login.
- No extra partitions mounted at boot.
The Results and Bootcharts
So the distributions I tried are openSUSE 10.3 Beta 1, Fedora 7, PCLinuxOS 2007, Kubuntu Tribe 4 and Mandriva 2008 Beta 1. What I found was that there was very little, if anything, between the boot up times of the various distributions. On my laptop however, openSUSE 10.3 Beta 1 came out as the fastest with the same impressive 27s!
Here is the run-down with bootcharts:
- openSUSE 10.3 Beta 1 in 27 seconds
- Fedora 7 in 41 seconds
- PCLinuxOS 2007 in 32 seconds
- Kubuntu Tribe 4 in 31 seconds
- Mandriva 2008 Beta 1 in 29 seconds
Now, though this is certainly an indication of boot time speed, boot time speeds can vary quite a bit on different hardware. There would be nothing impossible about different results on different hardware. That said, I would be very interested in seeing any comparisons on other hardware, so if you have some (from a default installation), please do share them as well.
In other exciting news, openSUSE 10.3 Beta 2 is scheduled to be released later today 
Edit: Digg, Beta 2 now out.
Powered by FireStats
Thanks for doing this (and sharing the result) – have you considered doing a similar comparison on login-times? Comparing the time from hitting after the password, until all applications have loaded on login?
One of the benefits with Linux is that you can keep your PC on 24/7, so login time is often just as important a measurement.
First-login can easily be tested with bootchart presuming you don’t have autologin enabled. But the first login and subsequent logins are completely different (second logins will always be a lot faster) because everything will be read from the disk’s cache instead of the disk itself, which avoids seeking. See Lunak’s Why does Linux need defragmenting? for more information.
Since openSUSE pushed the faster KDE start-up time improvements upstream there’s really not that much in it in terms of comparison of KDE start-up, which is why I didn’t really test it for each distribution.
[...] in minor tempo. Ecco i tempi di boot delle maggiori distro a confronto I dati li ho trovati su un sito, non ho potuto verificarli di persona. Purtroppo vengono paragonate soltanto le distro più [...]
you should try to boot this live distro
http://detaolb.sf.net (go get release 0.5)
it’s blazingly fast and small
christian
Arch linux is much faster
Thanks for the study. Can you please add Pardus 2007.2 (http://www.pardus.org.tr/eng/index.html) to the comparison? They also claim very fast bootup times.
Regards,
Fast
Arch Linux boots up within 10 seconds, with a fully configured KDE system and Beryl (note: more apps and daemons than the default of pre-packaged distros). Plus 5 seconds login time until everything has started, with 2GB system ram.
Dean: the point isn’t what’s installed but what you’re starting. Something tells me you don’t have KDM started in that time
. Note that I didn’t even use remap in any of the above bootcharts, since that would technically be cheating (the study is restricted to completely default installations). I’d be very interested in seeing a typical Arch Linux GRUB-to-KDM bootchart though, if you have one handy.
Come on arch fans you know as well as I do these aren’t fair comments. Arch is bare bones Suse is full blown.
Comparing a system thats a rolling release to one thats not, not to mention you maybe running a splitted kdemod which has been compiled against 4.2.1-3 compared to full blown kde.
Not to mention the possibilty of bakgrounding boot tasks
I have a pretty freaking powerful PC (it cost me 3400 USD) and last time I tried feisty it took about 40 seconds if not more to get to KDM… If you could share with me how you managed to get 31 I’ll be happy.
Serge, absolutely nothing special, like I said. It’s a completely default install with only bootchart added. Note that it was Kubuntu [Gutsy] Tribe 4 though, and not Feisty. I’m not sure if they’ve made any improvements during the releases.
One thing that can really increase boot time though is large, filled-up extra partitions, which take time to mount. Though of course if you have more services (i.e. http server, etc) starting it will also take longer. If you’re using reiserfs (as SUSE used to) it will also take longer
My laptop itself is ok but it’s really nothing special: 1.8 Ghz core duo and 1Gig of RAM. Needless to say, openSUSE at least boots a little faster on my desktop, which also isn’t _that_ great: 3400+ AMD64, 1Gig of RAM.
You reaaly need to repeat the test, making sure the same services are being called up, or at least list those louded by default for each distro.
Thanks for doing the test, it makes for an interesting read.
Matt: no, that’d be pretty useless. If you go down that road you might as well start making sure that each distribution has the same packages compiled in the same way as well, with similar patches etc.
The whole point of the test was to compare default boot times, since our classic Aunt Tille won’t be modifying any boot scripts or necessarily stopping and starting extra services. These tests and comparisons certainly are not immune from the “it runs faster but it doesn’t have functionality x, which Distro A has”. For example, Ubuntu doesn’t start sshd and smptd, which will speed up things a little but is also that little bit questionable.
That said, the nice thing about bootchart is that it will give you a nice representation of the services run and the disk utilization, which is why you can see where the problems are with delays in boot time.
Interesting, note that BeOS on my Celeron333 with 128Mo of RAM took 14s to boot from the bootloader to a fully functional desktop.
To do an equivalent comparison, you should have added the time to have KDE or Gnome starting in autologin mode..
So Linux has still quite a few progress to made on this point (and on the interface responsiveness).
[...] Giannaros has compared the boot times of various popular GNU/Linux distributions. Here is the run-down with [...]
I’m curious… Why all the fuss about something that only happens once during your PC operations. The longest boot time is 41 secs which is still shorter than M$ and it gives time to get coffee and then settle down to computing. I have a 3.2ghz HP with an Intel Dual core. Fastest thing I have had. Slowest was an 8080 running DOS… way back. But even then the boot was only once, at the beginning of operations. Then it was forgotten. Perhaps I am missing something. But I just don’t see the importance of booting up faster… unless it is life critical.
[...] read more | digg story [...]
Chuck. In the interest of saving electricity, I don’t have my PC on 24/7. I primarily use it for playing music or videos. A fast boot time is very useful….if you’re wanting to hear a song, it’s annoying to have to wait minutes for an OS to boot.
Ideally, I’d like an ultra fast booting OS that simply played mp3 and divx files, although this seems to only be offered as a commercial, sealed box product.
You should try source based distros like Gentoo or Slackware: once configured properly, they perform nice.
You should try source based distros as well: once configured properly, they perform nice.
Very well done issue about this good post! In fact, it is worth to buy an essay to receive an academic success!