Good articles comparing Linux and Windows performance?

I am doing a science project comparing the performance of Windows and Linux on sorting algorithms. I need articles about this written by people with standing (i.e. not students but people with a PhD).
Does anyone know of anyone with a high standing regarding this in the computer world or a good place to find articles? To qualify to submit to VJAS I need 3 print sources (i.e. books, encyclopedias, etc.) and 3 non-print sources (i.e. online articles) so anything would be helpful.

I have tried searching Google and I can find good articles comparing sorting algorithms and good articles comparing platforms, but can’t really find any using sorting algorithms to compare platforms.

On Wed, 02 Dec 2009 21:36:02 +0000, camper365 wrote:

> Does anyone know
> of anyone with a high standing regarding this in the computer world or a
> good place to find articles?

You might try checking to see if your school has back issues of articles
from the Association for Computing Machinery - you might find something
in there.

But in general, I would expect the impact of an operating system on the
execution of a sorting algorithm to be pretty negligible - sorting is
typically done in memory (unless it’s very large information and you’re
looking at performance benchmarking for disk I/O), and the OS generally
isn’t going to have a significant impact on the performance of an in-
memory sort.

Even then, you might broaden your search to look at comparative disk I/O
benchmarking in general as compared to search algorithm performance on
different operating systems.

Jim

Jim Henderson
openSUSE Forums Moderator

This does sound to me like a question set by someone who doesn’t know anything about sorting algorithms; it is normal to compare different sort algorithms against each other. As Jim says, how fast any one works will then be more dependent on the language used and the hardware available than on the OS.

Of course the simple answer is Linux because it is installed on the fastest supercomputer in the world.

So is there a good type of code to execute to compare the efficiency of different operating systems?

Hi,

If you are looking for articles written by people with standing, why not start with IEEE journal articles, or if you are at University try their online journal search facilities.

In my opinion these kinds of sources will be much more reliable than Google search results.

Regards,
Barry.

Have you read Sorting algorithm - Wikipedia, the free encyclopedia ?
This focuses more on the mathematics but may help you to determine which scholarly articles are more likely to answer your question.

You’ll have more luck comparing say BSD to Linux versions,
[Phoronix] Linux Hardware Reviews, Benchmarking, & Gaming](http://www.phoronix.com/scan.php?page=home) carry some benchmarks. They did attempt once comparison of Linux & Windows graphics driver once I think.

The problem is, MS designed Windows NT to be evil, deliberately encourage use of incompatible non-standard features, to maintain their application & user base lock in.

So what software can you run portably on Windows & Linux that does not give an unfair “native” advantage to one?

For example, suppose you test a windows program under Windows 7 and then in WINE? Is it a fair comparison.

On Fri, 04 Dec 2009 03:06:01 +0000, camper365 wrote:

> So is there a good type of code to execute to compare the efficiency of
> different operating systems?

Things that hit the kernel - for example, things that are heavy into I/O
would seem to me to be a good candidate for comparison.

But even then, I’d be inclined to do comparisons between like filesystems
(for example) - or raw I/O to and from the disk without regard for the
filesystem (IOW, using an unformatted disk and storing/reading data
directly to/from the disk).

Network I/O measurements using tools like netcat would probably also be
good, if you can find a version of netcat for Windows (it probably
exists, I just don’t use Windows so I never need to look for Windows
programs).

Jim

Jim Henderson
openSUSE Forums Moderator

On Fri, 04 Dec 2009 14:56:01 +0000, robopensuse wrote:

> For example, suppose you test a windows program under Windows 7 and then
> in WINE? Is it a fair comparison.

That’s not a bad comparison either; I noticed when playing Neverwinter
Nights years ago that running it under WINE was noticeably faster than
natively on Windows XP. I had two identical laptops (Dell C640s), one
with RedHat on it and one with WinXP on it - and the RedHat box ran much
faster with WINE and the Windows version of NWN than the WinXP box did.

Even then, I’d say it was perhaps not a fair comparison, because it’s
very possible the Windows box hadn’t been tuned to the degree the Linux
box had been.

Jim

Jim Henderson
openSUSE Forums Moderator

I write the code using standard C++ and use gcc to compile for Linux and mingw to compile for Windows. I’m not sure how much difference that would give, but I can’t really think of a better method.

I found that I/O is inconsistent within Linux itself. I found that tty1 (with the background image) produces output slower than tty2-6, which is slower than Konsole. How would that be compared with a Windows command prompt vs. command.com? (not that there’s much of a difference that I know of)

On Sat, 05 Dec 2009 01:26:01 +0000, camper365 wrote:

> hendersj;2081911 Wrote:
>> Things that hit the kernel - for example, things that are heavy into
>> I/O would seem to me to be a good candidate for comparison.
> I found that I/O is inconsistent within Linux itself. I found that tty1
> (with the background image) produces output slower than tty2-6, which is
> slower than Konsole. How would that be compared with a Windows command
> prompt vs. command.com? (not that there’s much of a difference that I
> know of)

I’m not talking about video responsiveness (what you’re seeing is that
tty1 (aka /dev/console) is a framebuffer device on most installations,
and tty2-6 are pure text consoles). I’m takling about raw data
throughput to storage devices and network controllers - that’s quite
different.

Apples and oranges to command.com vs. cmd.com (the latter is the 'Windows
command prompt" - and has significant differences from command.com); the
Linux analogy would be like bash vs. csh vs. tcsh vs. zsh vs. ash vs. …
(you get the idea).

Konsole and gnome-terminal just open a window to the default shell. You
get the same shell on the terminal console windows as well.

Jim

Jim


Jim Henderson
openSUSE Forums Moderator

Jim Henderson wrote:

> I noticed when playing Neverwinter Nights years ago that running it
> under WINE was noticeably faster than natively on Windows XP.

reminds me: circa 1993-?? (today?) OS/2 Warp ran Win3.1 in a “DOS
box”…and when it pulled a “blue screen” you just shut that “DOS box”
and launched a new instance of Win3.1, additionally things like DOOM,
WordPerfect for Windows, Lotus 1-2-3 and etc ran like the wind…some
of the Windows magazines noted that and started printing headlines
like: Warp is a better Windows than Windows.

which is when i think Bill and Steve decided to kill OS/2…

so, take care with what you say about the speed/stability of stuff in
WINE vs “the real thing”…(actually, i don’t think it matters, Bill
and Steve know it is already too late to kill Linux)


palladium

On Wed, 2009-12-02 at 21:36 +0000, camper365 wrote:
> I am doing a science project comparing the performance of Windows and
> Linux on sorting algorithms. I need articles about this written by

No article needed… there is ZERO difference.

Why would there be a difference?

I always thought the Linux community was polite.

The suggestion of IEEE articles was very good. Right now I am trying to apply as a member as a student. The abstracts of the articles look very good.

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

The GNU/Linux community is polite, the problem here is the question, the
algorithm performance is, in theory, the same, but in practical cases it
depends in how the operative system access the data, filesystems, memory
management, etc. In other words, yes, there is a difference, but isn’t
in the sort algorithm, the difference is in the OS design


VampirD

General Failure is the supreme commander of the Microsoft army.
All operation made by this army ends on him.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.12 (GNU/Linux)
Comment: Using GnuPG with SUSE - http://enigmail.mozdev.org/

iEYEARECAAYFAksmeAEACgkQuyH6KAqYAt2bxACdE2dIUurJQZAe2i5InF+/TqBL
pY0An3MlOgQGSO0YCSjhfLGJwW7qyOaA
=M6kn
-----END PGP SIGNATURE-----

On Mon, 2009-12-14 at 17:37 +0000, VampirD wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> The GNU/Linux community is polite, the problem here is the question, the
> algorithm performance is, in theory, the same, but in practical cases it
> depends in how the operative system access the data, filesystems, memory
> management, etc. In other words, yes, there is a difference, but isn’t
> in the sort algorithm, the difference is in the OS design

And in all fairness, based on the original question, this isn’t
going to matter… at least not much.

Which is why I said what I said. I thought it was succinct
and polite…

I found the comments fair. When ppl know the answers they want to hear, why don’t they just use search engines?

It was interesting to see what others thought about my idea of using WINE to compare. In past I’ve run Windows Peformance Index in a Virtual Box and found the disk subsystem actually works faster as a guest OS than when host! That rather suggests that Linux Block I/O kicks MS’s butt.