Fedora, Debian, FreeBSD, OpenBSD, OpenSolaris Benchmarks

[phoronix]

Phoronix: Fedora, Debian, FreeBSD, OpenBSD, OpenSolaris Benchmarks

Last week we published the first Debian GNU/kFreeBSD benchmarks that compared the 32-bit and 64-bit performance of this Debian port -- that straps the FreeBSD kernel underneath a Debian GNU user-land -- to Debian GNU/Linux. We have now extended that comparison to put many other operating systems in a direct performance comparison to these Debian GNU/Linux and Debian GNU/kFreeBSD snapshots of 6.0 Squeeze to Fedora 12, FreeBSD 7.2, FreeBSD 8.0, OpenBSD 4.6, and OpenSolaris 2009.06.

http://www.phoronix.com/vr.php?view=14533

[andrnils]

I just have a few thoughts that would make this test even more interesting, at least for me:

Setup each OS according to how they're normally set up. Out-of-the box benchmarks are only relevant for OS/distros that are run with out-of-the box binaries. For FreeBSD do a cvsup of both source and ports tree, and set some sane defaults in /etc/make.conf and rebuild world. A sane setup for the t61 is
CPUTYPE?=core2
CFLAGS=-O2 -pipe
I suggest also building all required apps from ports instead of using precompiled packages.
As Debian and Fedora is often run with out-of-the box binaries nothing needs to be done there. OpenSolaris benchmarks should be compiled with sunstudio-compiler suite, it's only a pkg install sunstudio12u1 away.

[vermaden]

ata(4) - no NCQ:
Bytes per second: 12455402 (11.8 MB/s)
Requests per second: 189

ahci(4) - with NCQ:
Bytes per second: 19889778 (19.0 MB/s)
Requests per second: 303

Enabling AHCI in FreeBSD 8.0 gives about 1.6 improovement in disk performance.

How to enable AHCI in FreeBSD 8.0:
ahci_load="YES" goes into /boot/loader.conf

Now you will need label your partitions and edit your /etc/fstab file like that:

1. label disks

root# glabel label root /dev/ad0s1a
root# glabel label swap /dev/ad0s1b
root# glabel label tmp  /dev/ad0s1d
root# glabel label var  /dev/ad0s1e
root# glabel label usr  /dev/ad0s1f

2. make changes in /etc/fstab

 #device		#mountpoint	#fs	#opts		#dump	#pass
+/dev/label/root	/		ufs	rw		1	1
+/dev/label/swap	none		swap	sw		0	0
+/dev/label/tmp		/tmp		ufs	rw		2	2
+/dev/label/var		/var		ufs	rw		2	2
+/dev/label/usr		/usr		ufs	rw		2	2
-/dev/ad0s1a		/		ufs	rw		1	1
-/dev/ad0s1b		none		swap	sw		0	0
-/dev/ad0s1d		/tmp		ufs	rw		2	2
-/dev/ad0s1e		/var		ufs	rw		2	2
-/dev/ad0s1f		/usr		ufs	rw		2	2
+/dev/cd0		/cdrom		cd9660	ro,noauto	0	0
-/dev/acd0		/cdrom		cd9660	ro,noauto	0	0

Reboot and check disk/storage oriented benchmarks again.

Disk/storage performance also depends on used filesystem, here are some rather old benchmarks of UFS/UFS+SU/UFS+GJ/ZFS/ZFS+COMPRESSION:
http://daemonforums.org/showthread.php?t=4264

[kraftman]

Good to see Fedora with SSE Linux enabled and Debian with old Linux kernel performed such great.

[malkavian]

I'd love to see comparisons that include Debian testing.

[glasen]

Setup each OS according to how they're normally set up. Out-of-the box benchmarks are only relevant for OS/distros that are run with out-of-the box binaries. [..]

The only problem here is that most of the benchmark tests of the PTS are compiled at runtime.

[Apopas]

Very interesting benchmarks.

Thanks a lot!

[DivineGrace]

I'd love to see comparisons that include Debian testing.

From the final paragraph of the article:

"If looking solely at the number of first place wins for each operating system, Fedora 12 and Debian GNU/Linux (2010-01-14) were tied with each having seven wins."

Since Debian testing and unstable are updated nearly everyday, wasn't either version dated 14-Jan-2010 compared with in this article?

[sabriah]

Nice test!

I can appreciate you choice to test just the official releases. Since we are talking about a myriad tweak options one has to settle on something and I think you've made a balanced choice here.

For each new parameter tested, like compiling options, version options, etc. we enter a permutation nightmare and prohibitively long running tests.

This is illustrative as it is.

Thanks!

[kebabbert]

I suggest you use the latest OpenSolaris patches and build, the next time you bench OpenSolaris. Just press "update manager"(?) and update to the latest build, which is build 131 right now. You are using build 111. There are new builds every other weeks.

One bench showed 76% higher performance just a few builds later. OpenSolaris is in heavy revision and has not hit Beta stage yet. It is in alfa stage right now, where new functionality is added. Performance is overlooked. Later, when all functionality has been added they will start to look at performance (and to iron out bugs). Just like game programming, only at the end they will concentrate on raising FPS. OpenSolaris is buggy right now.

The later the OpenSolaris build, the better it gets. So, I suggest you try the latest build next time. And when it hits Beta stage I bet it very fast.

I mean, to compare an OS in heavy revision, in alfa stage against a released OS is not really saying that much?