Heres a snip from the Gentoo weekly newsletter: Jose Alberto Suarez Lopez gave a presentation at HispaLinux 2003 where he demonstrated the load-time performance of the official Gentoo Linux 1.4 release. Gentoo Linux 1.4 for Pentium III, with and without prelink, were compared with a default Mandrake 9.1 installation on a Pentium III. The results – Gentoo Linux 1.4 with prelink did better than Mandrake 9.1 across the board, and even without prelinking Mozilla loaded nearly twice as quickly in Gentoo, and NetBeans loaded more than twice as fast.
Loading Benchmarks between Mandrake and Gentoo Linux
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Sounds like a biased test, imo. Some third party should do it, but unlike the previous Gentoo vs Mandrake and Debian artical, they should know what they’re doing with the Gentoo box.
I’m a full time Gentoo user, and I love it.
Yay for Gentoo.
…moving to source just has clear performance advantages at the cost of other issues like time, binary portability, ease of use, stability, etc.
It’s not like Gentoo is particularly awesome at building from source and does things any better than you or I can.
Gentoo didn’t invent source-based distribution, but it seems like they’re the only ones that get credit for it.
is it hard to include prelinking on mandrake? I first found out about prelink when i started using gentoo, and it was really easy to do. I cant imagine it being infeasible on mandrake.
On the other hand, apart from startup times being dramatically better or not, the overall feel of a gentoo system is a little smoother than other distro’s. I never really clocked any startup times, but I can really feel a (slightly) faster system. I have noticed this on other boxes too that went from redhat to gentoo.
…are these similarly configured Mozillas? The real test would be to build mozilla in /opt/mozilla from source in both distributions, and compare load times based on that. Mozilla can be built so many different ways that simple load times mean nothing by themselves.
From the article:
Mandrake 9.1 ships with its binaries “pre-linked,” which allows for faster load-times. In contrast, Gentoo Linux ships without pre-linked binaries by default.
No they are indeed not the first source based distro, but AFAIK they have the easiest environment, the largest package selection, the best docs and the finest userbase on irc and such.
I think they are getting credit maybe because they made source based accesible, not because they invented it
*Shrug* maybe you’re right. Honestly though, the more you learn about how to manage a Linux system properly, the less the actual distro matters. Gentoo users ’emerging’ around forums plugging the advantages of portage and building from source only reveal that they don’t understand that you can have the same amount of control with a RedHat system. In essence, in an attempt to show off “elite” skills, they admit to being novices.
But I don’t want to get down on other members of the community too much. Gentoo is a really, really good platform and has a great dev team.
Gentoo users ’emerging’ around forums plugging the advantages of portage and building from source only reveal that they don’t understand that you can have the same amount of control with a RedHat system.
The fact that a source based distro has those kind of users is a testament to dabooty is saying above. Gentoo brings the advantages of source to the masses. Also, you have to remember that those users are going to be the most vocal.
There are plenty of us who use Gentoo that don’t bother pushing it around forums. Because we know what you’re saying, that most optimizations can be done on any distro. Gentoo just happens to be what I prefer. But I have no problem pulling out a Slack or Debian disk when I have a different use.
-b
You’re rather mistaken. Its not so much the level of control, but how easy that level of control is. In Gentoo, I can update the ebuild for a particular software package very simply, and have the updated package installed onto my system within the package management framework. With RedHat, I can compile my own software, but if I want to install it within the framework of the package manager, I have to go to the trouble of making RPMs. Also, there are lots of little things that make Gentoo so wonderful. For example, the ebuilds for the NVIDIA drivers automatically patch the driver if it detects you’re running a development kernel. ccache and distcc are automatically integrated with Portage. The whole /etc/conf.d and /etc/env.d systems are very well designed. Packages that have additional setup requirements will usually display a nice little info message when the .ebuild finishes. Lastly, the community is great. The forums are *extremely* helpful, and the maintainers are very quick with new package releases.
… the best thing about Gentoo is the users, the developers, and the forums.
http://www.gentoo.org
http://forums.gentoo.org
@Vecc
I’m assuming OSNews thinks readers should at have at least half a brain to determine the quality/bias of a particular article. Relax and take it with a grain of salt. If you don’t think the article is fair because Gentoo conducted it, install and run Mandrake 9.1 and Gentoo 1.4 and see for yourself. If you have ever run either, you would know that Gentoo is generally faster than Mandrake. Doesn’t mean Mandrake isn’t fast enough or inadequate, I’ve had both running on three different systems and they both were great, but Gentoo has always been faster for me.
Errr…i have both of them installed. Thats not the point!!
What iritates me is that this is considered a news item, despite the fact that it’s from Gentoo ppl.
Btw, the fastest Linux system I have used is Vector! (Currently using Libranet).
-Vecc
What iritates me is that this is considered a news item, despite the fact that it’s from Gentoo ppl.
If you don’t believe them, why not try it yourself? It wouldn’t be hard at all to duplicate the test. If you then find the Gentoo study to be wrong for some reason, then post it here. Otherwise, quit whining.
I think people are too used to hearing performance reports from mega-corporations. Just because this study wasn’t conducted by an uninvolved third-party, doesn’t mean it is necessarily biased.
Personally, I think it is an attempt to honestly and accurately show the difference in performance between the two distrobutions. The fact that it was conducted by the Gentoo folks is irrevelant to me. At least they were honest enough to admit they did the test. Would you prefer them paying a supposed “third-party” to find the results like most OSes do?
Read again…..
“Errr…i have both of them installed. Thats not the point!!”
WHY DO PPL REPLY WITHOUT READING??? Amazing really ;:/
OK, slap me with a trout, i only once tried Gentoo around 1.0 and noticed they use devfs. I don’t like devfs, so i went away. Therefore i never ever really set up a Gentoo box and might understand the whole process wrongly.
If “prebuilt” means “precompiled”, i guess my point is moot, but after reading this comparison i wondered just how long does it take on a P3-733 to compile Mozilla? I have reports that it takes about 50 minutes on an Athlon 850. Let’s just asume that the P3-733 is just as fast. So we have 3000 seconds of compile time. For simplicity’s sake let’s say Gentoo started Mozilla 3 seconds faster than Mandrake. Easy to see that i have to start Mozilla a 1000 times to make up the time i lost during compilation. If i start Mozilla 5 times a day (which i usually don’t) i have won back the time i lost in compilation after 200 days. Hmh.
I bet that there is a new version of Mozilla available in the next 200 days and i if i want to update, i can never break even.
Maybe i am wrong and i am sure if so, it will be pointed out. But taking this into consideration i can’t see the enourmous advantage Mr. Lopez makes out.
And still: it’s only faster than Mandrake not “other Linux distributions”. I guess the Crux, LSF or Vector crowd will be very upset 😉
But it’s all about perceived time advantage….
If your compiles run while your computer is idle, you really lose no time (and it’s easy enough to make it update things overnight [or while you’re at work/school, etc]). Starting programs however is usually something that you do sit and wait for.
You’re rather mistaken. Its not so much the level of control, but how easy that level of control is. In Gentoo, I can update the ebuild for a particular software package very simply, and have the updated package installed onto my system within the package management framework.
You’re mistaken here. Building an RPM really isn’t much different from an e-build. Go ahead, poke around the documentation. Building a package is essentially the same in any system: tell it where the tarball is, tell it how to build it, and tell it what to do to integrate it into the rest of the system.
With a ports system, you have the spec files automatically downloaded in the ports tree. In RedHat, you have to either create them yourself or download them. That takes away a step in Gentoo that you’d have to do in RedHat, but doesn’t make the equivalent steps any more complicated in RedHat.
With RedHat, I can compile my own software, but if I want to install it within the framework of the package manager, I have to go to the trouble of making RPMs.
Honestly, that’s just ignorance. If I’m in Gentoo, and I download foo-1.0.tar.gz, untar it, ./configure && make && su -c make install, I haven’t gotten it to integrate with my wonderful magic amazing portage system. emerge -u foo won’t upgrade it; I would need to use Gentoo’s package system to take care of that. Building one’s own package in Gentoo requires the same basic skills as in RedHat–Gentoo is just set up a different way.
Also, there are lots of little things that make Gentoo so wonderful. For example, the ebuilds for the NVIDIA drivers automatically patch the driver if it detects you’re running a development kernel. ccache and distcc are automatically integrated with Portage.
I agree…gentoo has the best community support and is probably the best maintained distro out there right now.
The whole /etc/conf.d and /etc/env.d systems are very well designed. Packages that have additional setup requirements will usually display a nice little info message when the .ebuild finishes.
I haven’t researched this too much…my own system configurations usually involve two bootscripts, but then again I would never install a production server that way. If that is a great system, then I applaude the developers who came up with it, but no one is ever locked into any system. Even with such complicated massive platforms such as RedHat.
Lastly, the community is great. The forums are *extremely* helpful, and the maintainers are very quick with new package releases.
I’m not saying Gentoo isn’t cool. I’m just trying to get people to realize that it’s real merit is in it’s ability to give the really cool features that seasoned admins have available to less saavy users, not in this new revolutionary force that makes anything else a thing of the past.
>>But it’s all about perceived time advantage…. <<
I guess that you are right in a way. I am not convinced but for now i will accept the higher importance of perceived performance contra overall time consumption.
And to answermy own question: “prebuilt” apparently means “precompiled” and i wonder what happened to Gentoo. Are they something like Debian x86 on steroids now? I thought they were a source only distro.
The problem with this study is the absolutly wrong conclusion. The conclusion that the more optimized compiler flags of Gentoo enhence startup-time for the applications was not proven by this study. Not at all. For this you have to compare same configurations.
Different configurations influence the startup time of applications far more than any compiler flags can. (And we all agree that Mandrake’s and Gentoo’s default installation for say kmail are not the same.) For example I can make Mozilla’s startup time much longer by installing more plugins and doing nasty things with my font installation.
To say compiler optimizations make apps loading faster the test should compare Gentoo with Gentoo and vary compiler flags but not the configuration of the apps and their dependencies.
The test says startup time on Gentoo is better for these three applications then on Mandrake. But the test doesn’t tell us if compiler flags are the cause for this. I doubt.
>>>>But it’s all about perceived time advantage…. <<
>>I guess that you are right in a way. I am not convinced but for now i will accept the higher importance of perceived performance contra overall time consumption.
I’ve found that giving Gentoo an emerge -u world just before I go to bed doesn’t cost me any time that I’d be using otherwise, but I still end up with a current system in the morning. There are tons of cycles that your computer is wasting right now, probably, and you’re using it, right? Don’t assume that just becuase you can multitask, it means that you have to.
>>And to answermy own question: “prebuilt” apparently means “precompiled” and i wonder what happened to Gentoo. Are they something like Debian x86 on steroids now? I thought they were a source only distro.
The Gentoo maintainers did have *some* level of mercy on its userbase, and make a number of huge packages available as option binaries, usually named something like foo-bin, if foo is the source package. Gnome, OpenOffice, a Mozilla come to mind, and I believe there are a few others. I’m glad they did, too, otherwise I’d probably *still* be compiling KDE…
And all that being said, I’m very happy with Gentoo. I don’t feel l33t using it, or compiling it, or whatever, but I do feel in control, which is very important to me. No service runs on Gentoo with me specifically emerging it. No ports are open unless I tweak a config to make it so. No sneaky applications bloat up my hard drive unless they are dependencies for the applications I really want. Other distros behave this way too, I’m sure, but the point is the Gentoo behaves this way by deault, and easily, and I’m happy with it.
I mean, choice, right? What it’s all about? Picagentoo, I choose you.
I remember a similar comparison between load times for Mandrake, Gentoo and Debian, here:
http://articles.linmagau.org/modules.php?op=modload&name=Sections&f…
In that comparison Mandrake won easily, now in the Gentoo newsletter Gentoo win.
I don’t care really about who is the winner, but my problem is: both benchmark contradicts each other, I’d like to know why!
Maybe the hardware difference can explain it?
On one hand a P3 1GHz 128MB on the other hand a Celeron 2GHz 256MB, maybe the additional memory have helped the prelinked approach Mandrake, here..
Or maybe the benchmark differs from app to app, and if you pick your set of test you can show Mandrake or Gentoo as being superior?
This would mean that this “benchmark” interest is questionable at best..
> Building an RPM really isn’t much different from an e-build.
you mean e.g. “emerge openssh”
> Go ahead, poke around the documentation.
no – never.
>Building a package is essentially the same in any system: …
except with “emerge” on Gentoo.
> With a ports system, you have the spec files automatically downloaded in the ports tree.
… and apropriate config-flags, prefixes – the whole dependecy-packages inclusive.
> In RedHat, you have to either create them yourself or download them.
> That takes away a step in Gentoo that you’d have to do in >RedHat, but doesn’t make the equivalent steps any more >complicated in RedHat.
no – in Gento its is only _one_ step (ignoring the “emerge rsync”):
emerge PACKAGENAME
In Gentoo the whole download, read-the-README, configure,
post install depencies, make, make install thing is all done by the admirable Gentoo-package-maintainers and automated by “emerge”.
You haven’t used Gentoo before, have you?
And all that being said, I’m very happy with Gentoo. I don’t feel l33t using it, or compiling it, or whatever, but I do feel in control, which is very important to me. No service runs on Gentoo with me specifically emerging it.
No distro takes away that control, you just have to know how they work. Most binary based distros build from the top down…present a big Windows like desktop and force you to reverse changes they make to the default setup. Gentoo is built from the bottom up, and any time you move from top down to bottum up, you’re going to be able to skip a lot of hoops and get control. I’m just saying that the Gentoo team hasn’t done anything exceedingly groundbreaking; if you want to thank them for something, thank them for making it possible to tailor your system the way you want it sooner along in your Linux career than you would’ve otherwise.
3 seconds of time that you save on Moz startup if your productive time. You start it and wait for window to appear. 3000 seconds of compile time are not affecting your ability your ability to use the machine – you can compile in background or simply start the emerge when you go to bed. So you waste some CPU time and get more personal time in exchange. I don’t really mind if my CPU would be loaded a bnit more, it’s designed for that.
He was talking about downloading a source package and compiling, not using portage… There ARE programs out there that aren’t in the portage tree (like jahshaka from a previous news item here…)
And still: it’s only faster than Mandrake not “other Linux distributions”. I guess the Crux, LSF or Vector crowd will be very upset 😉
I never tried Crux or Vector, but while LSF might be faster, it takes much more time to make it functional. Okay, you could always use ALFS, but it removes the whole purpose of building its own LFS system. It would also be quite harder to manage, too.
I’m a Gentoo user, but it’s not the compilation from source that really attracted me, but rather the layout and the portage system. IMO they still need some work to do with that system, especially with unmerging (you just can’t unmerge a package and all its dependencies when you don’t want them anymore and emerge clean/prune are still a bit buggy), but it’s my favorite system so far. Being faster is a bonus to me as I would still use it even if it was a bit slower. Debian is not bad but I can’t get used to their default setup…
>> Building an RPM really isn’t much different from an e-build.
>you mean e.g. “emerge openssh”
No, there’s a difference between creating a package and installing a package. MY point was about creating your own packages, not how they’re installed.
>> Go ahead, poke around the documentation.
>no – never.
>>Building a package is essentially the same in any system: …
>except with “emerge” on Gentoo.
I guess I didn’t make myself clear. Again, I’m talking about building a package, not installing it. By ‘building,’ I mean creating a spec file or ebuild so the package is ready to be downloaded and installed. RedHat does the same thing as Gentoo, but it goes one step further and creates a package. MY point was that going from a source tarball to package that’s ready to be downloaded and built is the same on both systems, so you can acheive the same ends on RedHat and Gentoo.
The .rpm system isn’t limited in any meaningful way that the Gentoo system isn’t. But wait! What is this newb talking about? He’s never used Gentoo OR RedHat! Helllooo! RPM dependency hell!
Wrong; dependency hell is a by-product of binary distributions mixed with frequently changing software. The OSS community moves so fast that you either need a ports or apt-get system, or you need to avoid upgrading. But you can still build a RedHat box based on sources, and even avoid binary packages alltogether for the most part. People seem to think that rpms are difficult, but MY point was writing a gentoo e-build and building an srpm are basically the same thing. You didn’t address creating the e-build itself, because other people take care of that for you, and update the portage trees. But the real Gentoo users that manage ports have the same job as the guys that manage SRPMs.
I really hope you understand that, because if you don’t then there’s really no way to bring you up to speed here.
>> With a ports system, you have the spec files automatically downloaded in the ports tree.
>… and apropriate config-flags, prefixes – the whole dependecy-packages inclusive.
Actually, when you build rpms, you collect the same stuff–the same ./configure script builds the package. Of course, your claim that building packages for two different distros is radically different isn’t based on any fact.
>> In RedHat, you have to either create them yourself or >download them.
>> That takes away a step in Gentoo that you’d have to do in >>RedHat, but doesn’t make the equivalent steps any more >>complicated in RedHat.
>no – in Gento its is only _one_ step (ignoring the “emerge >rsync”):
>emerge PACKAGENAME
>In Gentoo the whole download, read-the-README, configure,
>post install depencies, make, make install thing is all >done by the admirable Gentoo-package-maintainers and >automated by “emerge”.
>You haven’t used Gentoo before, have you?
Ok, one last chance to jump about. I’m not talking about how the end user messes with his system. I’m talking about how the power user tweaks his system. I’m talking about the part of Gentoo only the devs deal with. It takes a lot of work keeping that portage tree stable and updated, and I’m sure they’d appreciate it if I enlightened you as to what exactly they’re doing for you.
They’re doing the same thing any other distro does. They’re maintaining a set of files that describe how to build and install the myriad of software that makes up your computer. When you actually learn to tailor your system, you’ll have to be able to do this at some point. Then you’ll come to realize that the kind of control that Gentoo gives you is not taken away by any other package management system. Gentoo doesn’t ‘put you in control,’ it just makes it easy to build from the bottom up.
> Honestly, that’s just ignorance. If I’m in Gentoo, and I
> download foo-1.0.tar.gz, untar it, ./configure && make && su
> -c make install, I haven’t gotten it to integrate with my
> wonderful magic amazing portage system. emerge -u foo won’t
> upgrade it; I would need to use Gentoo’s package system to
> take care of that. Building one’s own package in Gentoo
> requires the same basic skills as in RedHat–Gentoo is just
> set up a different way.
it is possible to “inject” foo into the wonderful magic amazing portage system. just do this:
$ emerge inject foo
and then
$ emerge -u foo
will upgrade it (if there is a version above of the installed version)….
For most rpms, you can copy a basic template specfile and only modify it slightly. Plus, some packages just have wierd build procedures, or require patches to work. It takes research to maintain packages, and I’m sure the portage maintainers are fully comfortable wading through e-build files.
Hmh, i try to buy the “perceived performance increase is more important than overall time consumption”, but i still have some problems with it.
Just some thoughts:
My PC at work (and all other boxes except the servers) are shut down each night. There are no “spare cycles” to do the compilation. Apparently Gentoo (or any source based distro) is not supposed to be used in businuess environments.
My home PC is a different matter, but i tend to switch it off. After all it consumes energy, which is expensive. But suppose i leave it on, how long does an initial setup take? How many programs are updated regularly? Do the updates sometimes require human input? Any info?
If it is anything like Debian SID (and the report claims as much “we try […] to offer the latest and best free software technologies”) i expect nightly downloads depending on your installed software from 10 to 100 MB in binary form. Don’t know how much this is in source, but i guess more 😉
Anyway, i guess for home users a nightly update will be OK, as long as huge projects like Open Office or KDE aren’t among them. Somebody on this thread already mentioned that he is grateful for binary packages of KDE from Gentoo. I fail to see the feasibility of the “perceived performance increase” when you have to deal with a one or two day long build process for your DE 😉
Whatever, the choice is yours.
The test struck me as a little unfair. Next time, try it with a normal kernel, instead of with the super-fast, super-unstable gaming kernel.
Also, versions weren’t mentioned. If Mandrake is using a different version of Moz, it could easily explain the difference in test results.
They also overstated many of the results – .4 seconds just isn’t much worse than .03.
-Erwos
In an office network, you might use distcc for distributed compiles. It can be run when there is not much activity around, and provide a hefty compilation speedup. But yes, you are right, Gentoo is definitely not aimed at business environment.
If you compile on your home box, just typing “emerge -u world” (to update all your apps) is enough. You can sleep, and everything would be done automatically. If you don’t want to upgrade everything, emerge -u app1 app2 app3 also works. Small stuff compiles fast, KDE or Gnome would probably take up to 8 hours (but not two days unless your machine is way too slow to actually run KDE
). However, updates of KDE and Gnome are not that frequent.
You’re mistaken here. Building an RPM really isn’t much different from an e-build. Go ahead, poke around the documentation. Building a package is essentially the same in any system: tell it where the tarball is, tell it how to build it, and tell it what to do to integrate it into the rest of the system.
>>>>>>>>>>
You don’t have to do that with ebuilds. Usually, you just copy an existing one and change the version number. That suffices for 90% of my uses anyway.
I would need to use Gentoo’s package system to take care of that. Building one’s own package in Gentoo requires the same basic skills as in RedHat–Gentoo is just set up a different way.
>>>>>>>>>>
The thing with Portage is that it usually trivially easy to make the package, so there is no reason to do the compile manually.
I haven’t researched this too much…my own system configurations usually involve two bootscripts, but then again I would never install a production server that way. If that is a great system, then I applaude the developers who came up with it, but no one is ever locked into any system. Even with such complicated massive platforms such as RedHat.
>>>>>>>>>
Its true that you’re not locked into any of the systems, but usually, if you want to use the system in a standard manner, you have to keep the standard systems in place. That standard system is much more palatable in Gentoo than in RedHat. You can go your own way, of course, but then you miss out on a lot of the nice work the Gentoo packagers do. For example, when I installed SpamAssassin, I was pleased to find out that the ebuild already installed a startup script for spamd in the rc folder. If I was rolling my own init system, I wouldn’t benifet from that.
You don’t have to do that with ebuilds. Usually, you just copy an existing one and change the version number. That suffices for 90% of my uses anyway.
The same can be said for SPEC files, friend. The point isn’t how easy or hard each distro makes it (Gentoo makes it infinately easier), the point is that the common perception that Gentoo is any more “robust” than any other distribution is just not true. The “robustness” of a system is entirely dependent upon the system administer.
The thing with Portage is that it usually trivially easy to make the package, so there is no reason to do the compile manually.
Yeah, true. And it’s best to choose a distro that suits how you manage your system. And Gentoo is a good distro for that. I just want y’all to to admit that anything you can do, anyone else can do too. And some of us don’t even need powerful programs like emerge to do it for us.
On Gentoo and techie forums, many Gentoo users give the impression that Gentoo’s package management system is superior to anyone elses. I’m saying at the fundamental level, they’re the same thing.
Apparently Gentoo (or any source based distro) is not supposed to be used in businuess environments.
>>>>>>>>>
You can always use the binary packages in the Gentoo Reference Platform (GRP). The real benifet of source-based systems is that packages don’t lag so far behind the latest releases. This is usually not a problem on a business desktop, where you have a stable, static configuration.
My home PC is a different matter, but i tend to switch it off. After all it consumes energy, which is expensive.
>>>>>>>>>
My PC generally does about 10 hours of compiling per week as a result of using Gentoo. However, I upgrade every week, and run KDE from CVS. A normal user will probably see 2 hours a week average. That’s not a big deal in comparison to the amount of time my PC is on, and when the machine isn’t compiling, it only takes up a few watts in sleep mode.
But suppose i leave it on, how long does an initial setup take? How many programs are updated regularly? Do the updates sometimes require human input? Any info?
>>>>>>>
Initial setup takes as long as you want it to. There is about 15 minutes of manual configuration to set up stuff like fstab and whatnot. Then, if you use the GRP, you can have a running system in another 30 minutes. If you compile, it can be anywhere from 6 hours (starting at stage2 and using GNOME) to 12 hours (starting at staage1 and using KDE). The only updates that require human input are binary packages that come from a server that has a click-thru license. This is stuff like Intel C++ or Sun JDK. These packages are clearly marked with a red ‘F’ when Portage gives you the summary of what it will install. Other packages don’t require human input by design, unless of course the compile fails. The latter happens occasionally, but then again I run KDE CVS and that’s KDE’s fault, not Gentoo’s.
Anyway, i guess for home users a nightly update will be OK, as long as huge projects like Open Office or KDE aren’t among them. Somebody on this thread already mentioned that he is grateful for binary packages of KDE from Gentoo. I fail to see the feasibility of the “perceived performance increase” when you have to deal with a one or two day long build process for your DE 😉
>>>>>>>>>
My DE’s build process takes about 8 hours once every few weeks (I track KDE CVS). I run it while I’m asleep, and its ready in the morning. Its really just a matter of suppressing the urge to have instant gratification, although you can usually rest easy in knowing that binary packages for software X won’t be out for RedHat for days.
NetBeans is a full blown Java IDE, taking up a *HUGE* amount of RAM.
So a few points:
* Were both systems using the same JDK (Sun/Blackdown/IBM)? Any difference may account for the loading of shared libraries. In addition Sun’s 1.4.2, the JVM isn’t mentioned in the article, is supposed to dramatically improve startup time.
* After you’ve loaded X there’s unsufficient RAM to do anything useful with an IDE such as NetBeans, particularly if Mandrake by default launches a number of services. The Mandrake version was probably swapping like mad! No one, I repeat, no one does serious Java development work under NetBeans with 128MB. Hence the load test does not represent real world conditions.
So, re-run the test with 512MB of RAM and ensured the same JDK is used!
I can tell you gentoo works just incredible as a headless server.
There is very little for a headless server that takes a long time to compile. It’s only the GUI stuff that takes a long time.
When SSH needs to be updated you just emerge it, and it a few minutes you have a custom compiled version perfectly matched to your server.
It takes about 35 to 40 minutes to install on a Compaq Dl360/380 using the live cd (complete with raid), and you can do the install remotely using SSH, how cool is that.
The gentoo gaming kernel is not unstable, where did you pull that one from? I use it on my Desktop/Server and it is rock solid.
Gentoo can work in a business environment, have you heard of Ghost or similar, you create one image and then ghost it to other PCs. Most business standardize on common PCs, so it’s not as big a deal as you suggest.
Also something of interest in this test, after the browser was opened users spent an average of 2.5 seconds trying to remember why they opened it, 6.2 seconds typing in the URL and 48.9 seconds day dreaming about that hot blond in cubicle 4A. Gentoo users responded by saying that’s because they didn’t compile with the –nosex flag.
“The gentoo gaming kernel is not unstable, where did you pull that one from? I use it on my Desktop/Server and it is rock solid.”
Your sample space of 1 means what? (hint: nothing)
Pre-empt and low-latency have _known issues_. They are not, repeat, not suitable for serious use at this time. The fact that they don’t crash for you only tells me that they don’t crash for you. If they were entirely stable, don’t you think they would have been put into the regular kernel by now?
-Erwos
//Pre-empt and low-latency have _known issues_.//
I have a low-latency and premptible kernel. No issues here. So make that two. In fact many Gentoo users use low-latency and preemption. It hasn’t crashed on me in the 7 months I’ve used it. Red Hat on the other hand, not to mention Mandrake, have forced me to restart on occaision. I still have a Red Hat box but only because it is so simple to set up. I really only use my Gentoo laptop though because no matter what everyone wants to say, it does give you control. Obviously you can have the same amount of control in any other distro, they’re all open source after all, BUT Gentoo gives you the tools to do it. USE variables give you much more control over what support gets compiled in without having to edit RPM spec files individually. SLOT is another great tool, allowing you to have multiple versions of the same program. The init scripts are a huge improvement over sys V and rc-update makes things even better. Gentoo is more than a sum of its parts though. There are many things that make it a wonderful distro but the whole package is what really sells itself to the masses.
The problem here is perception. Most Gentoo bashers fight about the apparent speed increase while most actual Gentoo users use Gentoo for reasons other than speed. With that said, Gentoo has definitely been faster for me than Red Hat. Could Red Hat be tweaked to make it faster than my Gentoo install? Maybe, but that’s not the point. Gentoo makes it easy for newbies to use a source distro but at the same time it is extremely powerful in the hands of an experienced user. That may be a reason it is so popular.
Amongst newbies it is popular because of its friendliness towards them. Unlike some distros (cough-debian-cough), people don’t shut you out when you’re clueless. I felt bad for newbies trying debian. I was one of them years ago. It seems to be even worse now but it wan’t too great years ago either. It’s never been that way in my experience with Gentoo. The forums are filled with helpful people willing to get someone through their first install. I rather enjoy helping people when I can.
The other side of Gentoo is the power users. You’ll find people integrating portage with BSD, porting Gentoo to other architectures, and using Gentoo for embedded devices. These are not tasks left to newbies. The intelligent people working on these projects often chat side by side with normal users, even newbies with nary a cruel remark.
Who cares about a 1ms vs 2ms load time on a P4 3 Ghz.. however you would like to have a nicely optimized OS/Desktop on machine with limited ressources (old cpu, etc) but those are precisely computers the less likely to compile all the stuff needed
I have a PII 400 192Mo with Mandrake 9.1/Gnome, it’s not perfect but speed is correct. Maybe a Gentoo would do better but I fear the compilation process on this machine will take forever (and I probably don’t even have a partition big enough to compile Gnome..)
That it’s biased by the source is irrelevant. Even if methodologically correct, the problem is that the conclusions are nonsensical.
Things work faster, therefore it’s because the code is faster. This is almost never true in practice. Performance, especially in things like application load times is almost always dominated by something other than the performance of the raw code. I/O is usually the most significant factor.
Was the test performed using two identical machines (or one machine at different times), using the same filesystem, using the same partition layout (hard disk performance varies by tens of percent between the beginning and the end)? Even if everything was “the same”, there are a lot of secondary issues that can have effects on the performance (e.g. the order of package installation affects the relative placement of the main programs and their shared libraries).
Benchmarking should really require a license.
What shocks me is that the Gentoo project would post something like this on their page. It certainly lowers my opinion of them.
Is there any Linux distribution that is flexible, up-to-date and governed by knowledgeable people?
http://archives.mandrakelinux.com/cooker/2003-10/msg00758.php
There is no prelink in any Mandrake Linux distribution yet, according to Gwenole Beauschene on the Mandrake developper mailing list (cooker), and he points out other inconcistencies. Apparently the author should make his homework and give more information about how the data were measured before being trusted.
billybob: >> Gentoo can work in a business environment, have you heard of Ghost or similar, you create one image and then ghost it to other PCs. Most business standardize on common PCs, so it’s not as big a deal as you suggest.<<
My company runs a very diverse set of PCs, ranging from original Pentium-MMX 166 MHz (Terminal duty to a partner’s data warehouse) to Intel P4 or AMD Athlon XP for graphics or data processing. Let’s call it a grown structure and for us it’s a big deal.
billybob: >>It takes about 35 to 40 minutes to install on a Compaq Dl360/380 using the live cd (complete with raid), and you can do the install remotely using SSH, how cool is that.<<
Is this with or without the configuration of the services offered by the server? And, i didn’t know that there is any other way to set up a 19″ rack server, but over the net. The basics, that means without the specific service(s), take about 15 minutes with our preferred distro.
Rayiner Hashem: >>Its really just a matter of suppressing the urge to have instant gratification, although you can usually rest easy in knowing that binary packages for software X won’t be out for RedHat for days.<<
I don’t care about Red Hat. That said, i thought that the “perceived performance increase” is so important. So when starting an application it is important to have a reaction 3 seconds faster, but when i install new software i can wait 8 hours? Excuse me? Ahh, probably it is just beyond me, i am sorry.
The choice is yours.
Hi all,
I personally know the guy that made this benchmark (not Jose Alberto, he just made it “public”). There was not intention in making it “public” nor much detail to make this a “really ideal test”, but heck, others dont do either.
This benchmark was done using a kind of old computer the developer had at home (his girlfriend was using it with mandrake). As Jose Alberto convinced him to switch to gentoo because all our company computers are bieng ported to it, the developer decided to run some “home made” tests with mandrake (just a moth old) and the just installed Gentoo.
Of course this computer was not being used for “real work”, just surfing and playing KDE games
But Jose Alberto made the results public in last Hispalinux. This wasnt either a “serious effort” it was just to poke a little bit all Debian users in the room
What has surprised us is the relevance the benchmark has gotten. First it appeared in Gentoo main page and now in osnews, the only thing that lacks to make this even worse would be to appear in Slashdot
)) Actually, we received an email from Mandrake noticing that Mandrake doesnt use prelink binaries as Juan Jesus (the developer) thought.
So this was just a “home made” test, nothing to convince your sales staff nor CEO, just some playing around.
Hope this clarifies the situation. If we have the time, we WILL run a more fair benchmark, giving more details to all of you, but next time Microsoft will pay for it and it will show Windows XP is muuuch faster than Gentoo and Mandrake
Regards.
I have both systems installed and my Mandrake is faster, but I love the way Gentoo handles pkg management. I think it is an apples and oranges problem. I would really like to find out why my Gentoo install is slower (I used the 1.4 p3 cd) on a P3 system, but for some reason the Mandrake is faster at opening programs and running X?
Pre-empt kernel unstable?? I think not. First of all most of those options are also available in gentoo-sources. I happen to have all of your so called unstable performance switches turned on on a server here at work, my server at home, and 2 desktops. All the systems are heavy work horses and havent batted an eye……Im talking 100% uptime. I recently upgraded my personal desktop to 2.6-test5 mm-sources and that machine has been running 24/7 for about 2 months now without issue.
Just because there are some (few) reported issues of pre-emp and the likes causing a some ppl problems doesnt mean its unstable. I also know RH9 shipped with pre-emp turned on by default too.
The prelink issue is really a moot point. Although this was a small quick test, it clearly shows that even with prelink disabled on Gentoo it still manages to load things nearly twice as fast. It all comes down to how you setup your system. What kernel youre using with what patches and which options and pre-compiled binaries or system specific source compile.
Im sorry, but no matter what, if you set Gentoo up right it should always be faster than the rest. I ephasize the “set it up right part” Ive seen Gentoo systems setup generically with no cpu specific compile flags and no performance enhancements in the kernel get easily beat by current RH, SuSE, and Mandrake. This test just shows that when you setup a Gentoo box the right way its faster.
I love Gentoo and I spent about a month trying to get everything running as fast as I could on my PII 300mhz pile of junk. I tried all the different optimization levels, all the CFLAGS, all the prelinking and on my machine, with the Gentoo gaming kernel, Gentoo was noticably slower. I don’t know why, I don’t know how, but different people on different boxes will experience different things. (And yes, I set up Gentoo “the right way”.)
I really love Gentoo and it’s packaging, but to be honest with you, I got tired of 65 hour KDE compiles on my box just to see KDE run at less than half the speed. (I don’t want to hear people saying that I shouldn’t be running KDE on this box to begin with. It’s been running quite well for me.) Once I get a faster box (PIII or above), I’ll give Gentoo another try because I really want to make Gentoo work for me. For the meantime though, Mandrake Cooker has made me more than happy in terms of speed and stability.
So… anyone got any spare hardware they could sling my way?
Oh my God,
Jose Alberto is receiving emails asking why did he publish the results and not Juan Jesus. OK, I’ll try again and hope this solves all the issues.
Jose Alberto (JA), Juan Jesus (JJ) and myself work in the same company. JA decided to test wether Gentoo was faster than Mandrake AT HOME because we read a time ago the benchmarks about mandrake, gentoo and debian. As JJ had a Mandrake computer at home, he made some test just for fun. They were not intended to go public, nor done with very much detail and care, just a quick test.Of course, JJ received all the needed help from JA as they are friends besides workmates.
At the same time, JA was preparing a presentation about Gentoo for Hispalinux. He decided to show the results just to poke a little all those Debian users in the crowd. This was NOT intended to go BEYOND THAT ROOM and moment. BUT this results got some attention from Gentoo guys, they made them WORLD public in their main website and everything is now rolling as a snowball.
As this has gotten such public attention we will try to repeat them with much more care to detail, but only when we have some spare time.
So, JJ and JA are just two friends that work as developers in the same company and decided to do some “home made benchmarking”. This was not intended to go public, specially not world wide public. As we all know, even bechmarks done by pros have biases and show “what they want”. Please, just cool off and take them as intended, just a game, not a public affair.
Hope whis clarifies all. Regards.