[PHORONIX] Debian 7.0 GNU/Linux vs. GNU/kFreeBSD Benchmarks

And another OS war has started in the comments, as usual.
Oh, this time I am / we are "mentally abnormal". Cool! :beergrin

(Well, at least I don't waste my time barking as an hysterical dog every time someone approaches my owner. Stay calm, doggie, that big ugly red daemon won't eat you. Keep gnawing your bone. Good doggie.)
 
Hahaha, check this out.

kFreeBSD ahead of Ubuntu and other distros for some tests. Interesting :D
 
From my understanding, the kFreeBSD uses the FreeBSD kernel and userland with a libc plus Debian and other applications. "The same userland" as stated by Phoronix is a false statement.
 
From what I've read on Debian GNU/kFreeBSD website, it's just the kernel that comes from FreeBSD; world is built from GNU sources (unless there is some program in the world, that is so different from the GNU version and is needed for the kernel and the system to operate safely) -- that's what the `GNU' stands for in the name.
 
sossego said:
From my understanding, the kFreeBSD uses the FreeBSD kernel and userland with a libc plus debian and other applications. "The same userland" as stated by Phoronix is a false statement.
Click to expand...

Yes, they use different userland.

I think they should compare FreeBSD and Linux kernel without any WM/X.Org at all so its fair.
 
alie said:
I think they should compare FreeBSD and Linux kernel without any GUI at all so its fair.
Click to expand...
Agreed, but I don't see that happening any time soon to be honest. The thing is; to get something like that set up you'll need to have quite some experience with both environments. But people who do usually don't care about these things at all because they already know when to use either Linux or FreeBSD, or Windows for that matter.
 
Maybe I am blind, or there has been some changes to the site - but I see no test listed where GNU/kFreeBSD was better.
 
One should pay mind to the 'less is better' and 'more is better' remarks in the graphs.
 
Crivens said:
Maybe I am blind, or there has been some changes to the site - but I see no test listed where GNU/kFreeBSD was better.
Click to expand...

He's right, I see no test listed there where FreeBSD came out on top. D'oh.
 
Slightly off topic (small rant!), but I somewhat hope no one will notice in here.. :e

I just skimmed the rest of that website and also discovered the forums and quite honestly it only made me appreciate this place a whole lot more. Not so much because of the topic at hand, but because of the way this forum is being run.

I honestly don't care what the overall topic of a forum is or if I agree with the overall opinion or not. I think it's a depressing sight for eyes if you almost can't open a single forum thread without several trolls being obviously present. Sometimes even only obvious for the trained eyes, I'm not too sure about all of them, but here's where the trolls get their way: with so much "noise" being present it becomes almost impossible to separate a troll from a critical poster.

Something which I think can really hurt your forum. Sure; things shouldn't get out of control, but there's also nothing wrong with a good (sometimes heated) debate between people, as long as they can show some mutual respect for each other. And obviously moderation shouldn't be overdone either way (* remark), but even so; loosening up on moderating posts like they seem to be doing can very easily make a whole forum a lot less appealing to many potential visitors. Especially if they get the feeling that they need to pay extra attention with every post they make as to who they're answering because up front there's no telling if you're feeding a troll or participating in a heated discussion.

I think that's really sad, because no matter what your personal opinion on a specific forum is; never forget that actually running and administrating one takes up a lot of your time. A lot more than sporadically visiting and sharing some comments every now and then.

* Remark: Although I'm commenting on moderation in general it's not my intention to start discussions on that matter with regards to this forum.
 
segfault said:
He's right, I see no test listed there where FreeBSD came out on top. D'oh.
Click to expand...

Well, first of all, it's not FreeBSD, it's Debian GNU/kFreeBSD.

The following may sound like a rant, but it in fact isn't. I am in a good mood an I am only writing this to try to motivate this community.

In my opinion, this can hardly be measured as a test of FreeBSD performance. For several reasons, most obvious are
  1. This test measures incompatibilities between the GNU userland and the FreeBSD kernel, and NOT FreeBSD performance. GNU userland has been with time honed and polished for work with the Linux kernel (or rather, the Linux kernel development has been steered such that it accommodated the GNU base system).
  2. If they wanted to check performance Debian vs. FreeBSD, they would've just used FreeBSD, and not Debian GNU/kFreeBSD. FreeBSD is an entire operating system in its own right, not like Linux, which is really just the kernel.
  3. As far as I have understood, this release of Debian GNU/kFreeBSD is more like a beta version, and they released it for testing, not for production. There are things that are likely to break, and as such, this test shows how far away the development of Debian GNU/kFreeBSD is from the Debian GNU/Linux.
  4. And of course, last but not least, performance-wise tests depend entirely on the system administrator. You can have a badly configured FreeBSD, or Linux, or you can have someone who really intimately knows what's going on under the hood.

There are some (important!) philosophical disagreements between Free Software Foundation, FreeBSD Foundation and the Open Source Initiative, but the end-user wants (in most cases) to see performance and quantitative facts to which one can hold on to. If you think you can do better, you can do your own tests for cross-checking, heck, why not just try installing benchmarks/phoronix-test-suite and see what gives (true, those are different tests, but nonetheless).

Maybe write an article, and send it to the BSD Magazine, about how those tests are biased, or give them praises for objectivity. There is so much more that you can do, instead of just saying "those things are rubbish!" and attacking them with suspicion instead of facts. The code is there. The experimental method states, that all experiments must be reproducible. Well, here's your chance.
 
Benchmarks are not real world situations when it comes to programming and system use. I set up systems for audio production and there are no benchmarks for such. System performance depends on a few more variables such as:
  • Memory
  • CPU type and possible subset
  • Whether or not a GUI is being used
  • The GUI being used
  • The libraries being used. I.e. Qt4/3, GTK2/3, FLTK, Python version, et cetera & et al
  • Kernel release, hertz rate
.
 
True, hardware plays an important role, however in such tests, they usually double check on the same hardware. The point was, that you can have a very good hardware and the latest kernel and so on, but if your system is poorly configured, an older and cheaper computer that has been properly configured and optimized for your use, can work considerably better.
 
The exact same system will perform differently on a PS3 and a Thinkpad.

Back to the "same" hardware. If the system(s) is optimized, then performance will be better.
Look at the list in my previous post. Every environmental variable will effect the performance. I'll stay with my opinion on benchmarks: They are not applicable in real world situations. These are no different than a wind tunnel test. The wind tunnel test will not tell you what happens when a tornado or hurricane strikes your immediate vicinity because they are in a controlled environment.
 
I'd like to see a three-way test between Debian's GNU/Linux, Debian's GNU/kFreeBSD and pure FreeBSD. It would be interesting to see how stock, out of the box installs of the three would do. If one is consistently slower, as we saw in the linked test, I'm curious as to where the bottleneck is. Is it hardware drivers, a file system, process scheduling? As someone who occasionally uses all three operating systems I'm less interested in how one or another fairs in any one test and more interested in __why__ one does better than another in a given test.
 
Instead of seeing out-of-the-box performance, I'd like to see both systems tuned properly, and see some tests of binary performance of both. Something like benchmarks on benchmarksgame, so memory usage comparison, user time, CPU time ...

Each test would be a numerically difficult problem to solve, both Linux and FreeBSD "camp" would optimize their systems as they want, and then run the program. Maybe this sounds like some sort of Olympiad, but what matters isn't out of the box experience, but tuned performance, for which you need both camps to collaborate/compete.
 
I have to disagree that tuned performance matters more than tuned for two reasons. A) Most people don't tune their systems and run with the defaults. B) Each test would require different tuning which would make the results only applicable for that specific test.

That being said, I do like the idea of getting groups of experts from two different platforms competing in a sport-style match off. Sort of like a decathlon for computers. Run both teams/OSes through a series of tests and see which one scores better in each "event" and overall. I think that would be good entertainment (and educational).
 
Not only are they using a X.0 kernel, which is notoriously known for having extra debug features which makes it slower, but the tested systems are also not comparable. Phoronix links to this benchmark as their source (end of third paragraph). Some things which stick out:
  • Motherboard: LENOVO 6459CTO vs LENOVO TP-7L
  • UI: Gnome Shell 3.4.2 vs Xfce 4.8
  • Display Driver: Noveau 1.0.1 vs VESA 2.3.1 (any reason they couldn't use the binary Nvidia driver?)

There are also several benchmarks which were only run on or shown for Debian Linux. Some of them: security/botam, multimedia/libvpx, parallell bzip compression, games/crafty, and audio/lame. Need I go on?

They don't seem to provide any reasons for why these and several other tests were not run on Debian GNU/kFreeBSD. This is obviously not an objective test, as the results could very well be cherry-picked.
 
Savagedlight said:
There are also several benchmarks which were only run on or shown for Debian Linux. Some of them: security/botam, multimedia/libvpx, parallell bzip compression, games/crafty, and audio/lame. Need I go on?
Click to expand...
No, I think not. IMHO we are at the point where the next logical question is: "Who paid the piper?"
Savagedlight said:
They don't seem to provide any reasons for why these and several other tests were not run on Debian GNU/kFreeBSD. This is obviously not an objective test, as the results could very well be cherry-picked.
Click to expand...

In scientific literature "Typical results are shown" means "the rest did not make any sense". With benchmarks (and politics), it means the rest made us look bad.
 
throAU said:
Performance on all is good enough that for 99.99% of users it is pretty much a case of "meh".

I want to run a Unix distribution. Linux is not Unix.
Click to expand...

To be quite honest, neither is FreeBSD anymore. See FreeBSD C99 & POSIX project. However, I really can't wait for the day, when I'll be able to say that I'm using FreeBSD Unix.
 
You must log in or register to reply here.
Back
Top