Some interesting benchmarks...

Gullible Jones · 2005-10-12 23:47:06

Yes, I know that benchmarks don't necessarily correlate with realtiy, but I find these quite interesting, despite being a bit dated. To summarize the results:

- Linux 2.6 performs extremely well.

- FreeBSD 5.1 performs very well, and pulls some rather strange stuff.

- Linux 2.4 performs surprisingly well.

- NetBSD performs well enough.

- OpenBSD performs poorly.

Now for a few comments:

- I think the bloke bashes OpenBSD a bit much. I like the "security by default" ideal, especially given the state of desktop-oriented Linux distros. I have heard, though, that OpenBSD has very slow I/O - has this been improved at all?

- I had no idea before that FreeBSD has a default limit on the number of processes. I like this - it could make fork bombs much less effective, and ought to be implemented in Linux. (If it hasn't been implemented already, that is - kernel development is so damn fast these days!)

- Anyone know the reasons FreeBSD's odd, sudden performance jumps after loads reached a certain point in some of those tests?

- Would this set of benchmarks be likely to show higher performance for Linux 2.6 using the current 2.6.13 kernel? 2.6.13.x seems faster than previous 2.6 kernels for some things...

Edit: Hmmm... it seems that newer versions of NetBSD scale much better. Interesting. I wonder if OpenBSD has improved similarly since the original benchmarks?

sepht · 2005-10-13 03:27:08

Ancient and old benchmarks man.. ANCIENT. 2002 or 2003. with the production release of 2.6, movement to new development, NetBSD 2.0, FreeBSD moving to 5 and 6 instead of 4...

Here's a set of MySQL-based benchmarks for FreeBSD, NetBSD, OpenBSD, Linux2.4, Linux2.6, Solaris, and all the little options for FreeBSD that it offers.
circa, feburary 2005, newsforge.

http://software.newsforge.com/article.p … =72&tid=29

Gullible Jones · 2005-10-13 09:33:15

Wow... I had no idea that the Linux kernel had gotten that good.

demonus · 2005-10-13 10:13:42

Gullible Jones wrote:

- I had no idea before that FreeBSD has a default limit on the number of processes. I like this - it could make fork bombs much less effective, and ought to be implemented in Linux. (If it hasn't been implemented already, that is - kernel development is so damn fast these days!)

it is implemented but no limits are set as default, loot at /etc/security/limits.conf

Gullible Jones · 2005-10-13 10:33:26

Should be implemented as default then. I think a limit of several thousand processes would work for neutralizing fork bombs.

demonus · 2005-10-13 11:00:24

it's up to the distro maintainers

Snowman · 2005-10-13 18:33:36

demonus wrote:

Gullible Jones wrote:
- I had no idea before that FreeBSD has a default limit on the number of processes. I like this - it could make fork bombs much less effective, and ought to be implemented in Linux. (If it hasn't been implemented already, that is - kernel development is so damn fast these days!)
it is implemented but no limits are set as default, loot at /etc/security/limits.conf

There was a thread (discussion) about that a few months ago. A forum search will certainly find it.

Arch Linux

#1 2005-10-12 23:47:06

Some interesting benchmarks...

#2 2005-10-13 03:27:08

Re: Some interesting benchmarks...

#3 2005-10-13 09:33:15

Re: Some interesting benchmarks...

#4 2005-10-13 10:13:42

Re: Some interesting benchmarks...

#5 2005-10-13 10:33:26

Re: Some interesting benchmarks...

#6 2005-10-13 11:00:24

Re: Some interesting benchmarks...

#7 2005-10-13 18:33:36

Re: Some interesting benchmarks...

Board footer