Ubuntu vs. OpenSolaris vs. FreeBSD Benchmarks

[andrnils]

As far as I know ZFS port in FreeBSD is in terrible state (or was).

I'd say was. There has recently been a big ZFS related commit to -CURRENT. Some more info here

I can't imagine why someone didn't use stable version of FreeBSD in tests? In my opinion benchmarked systems should be tune for the best performance like in professional benchmarks. Default settings can be very different - Ubuntu is desktop distro and FreeBSD is probably tuned for servers etc.

No, it's a bit strange...

[kebabbert]

it would nice to see a linux that is not cursed with ext3

btw, if you care about filesystem performance, read this;

http://bulk.fefe.de/lk2006/bench.html

in short, solaris lies, bsd sucks.

If you learned to read an article well, you maybe wouldnt draw wrong conclusions too quick.


In the article, they tried 2 old OpenSolaris distros v0.56 by some random people. They didnt try the real SUN Solaris. So infering that "solaris does lie" is totally wrong. Maybe you could infer that "2 opensolaris distros v0.67 lies" - but, did the people behind the distros promise anything? No. So, wrong conclusion again.


And remember that it is well known that Solaris does not shine with one or 2 CPUs. Solaris shines when you have lots and lots of CPUs in large machines. Then Linux sucks, there are several links about Linux bad scaling when using many CPUs.


Sure, there are Linux clusters with many CPUs, but they are tailored to a specific task and can do nothing else. The clusters suck for ordinary OS usage. They can only crunch numbers and nothing else. Everything is ripped out from the Linuxkernel; drivers for webcams, etc.


Linux uses a naive approach that works good for desktops and small computers. You do know that Linux comes from the desktop and now tries to become a server OS? Solaris comes from the Server halls and now tries to become desktop. To use lots of CPUs efficiently, you must use a complex and heavy design that takes many years to develop - and that machinery does work best in large servers. Linux would suck in one of these, because it scales to bad. Bad threading model, etc. There are lots of links, if you google a bit on that. But for simple tasks in large clusters, Linux is good. Easy to modify. Solaris kernel is difficult to understand, Solaris is from 1982, but was called SunOS then. It is very mature and rock solid. Complex. Mature.

[kraftman]

Why not throw in a test with FreeBSD installed on ZFS. There is native support for it since 7.0. I also agree that doing performance tests on betas and release candidates aren't that infomative. So a reiteration on 7.1 RELEASE and 2008.11 when they release that might change things.

I would also love to see a test of how they fair power management-wise.

As far as I know ZFS port in FreeBSD is in terrible state (or was). I can't imagine why someone didn't use stable version of FreeBSD in tests? In my opinion benchmarked systems should be tune for the best performance like in professional benchmarks. Default settings can be very different - Ubuntu is desktop distro and FreeBSD is probably tuned for servers etc.

[andrnils]

Why not throw in a test with FreeBSD installed on ZFS. There is native support for it since 7.0. I also agree that doing performance tests on betas and release candidates aren't that infomative. So a reiteration on 7.1 RELEASE and 2008.11 when they release that might change things.

I would also love to see a test of how they fair power management-wise.

[kraftman]

What? You have no idea who I am and I have no idea who you are. I registered because I've enjoyed the lot of phoronix articles...

Also, anyone who knows anything about gcc, knows that -g can affect performance, particularly with data structures like those used in DBs. Run a B-Tree in debug and "release" mode and you'll see what I mean.

Peace,
pcfxer

I know what you mean. I just replied in such tone, because I thought that your previous post was attempt to make flame (I shouldn't reply at all )

this one just seemed like it had a terrific idea but an inconclusive result.

Of course, this and previous one, but what can we do?

[energyman]

emm, in a 'fair' test jfs is beaten by everyone. Oh, and xfs? It is well known that its defaults suck. In its standard configuration it is dead slow - but with just a little bit of tweaking it does become A LOT faster. And I am not a xfs fan...

[curaga]

There's Linux file system which uses very low CPU too - JFS, but its performance is probably lower than EXT3 etc.

Actually JFS beats ext3 in almost all tests I've seen. It's a stable performer, unlike ext3 and XFS that are good in one thing but bad in something else.

[pcfxer]

Maybe your hardware is a problem, because I don't notice any responsiveness issue related to Linux kernel. Do you believe that such regression wouldn't be noticed for years?

@pcfxer

Fanboys comming... Are you dumb or something? Benchmark shows that Linux beat your FreeBSD... Idiotic attempt to make a flame war. If you're interested I can show you a benchmark in which Linux kernel 2.6.22 and newer one kick your lovely FreeBSD 7 or 7.1.xx.

And you registered to write such a bullshit? xd

EDIT:

@Bugmenot

I missed your post before. You're completely right!

What? You have no idea who I am and I have no idea who you are. I registered because I've enjoyed the lot of phoronix articles, however, this one just seemed like it had a terrific idea but an inconclusive result.

Also, anyone who knows anything about gcc, knows that -g can affect performance, particularly with data structures like those used in DBs. Run a B-Tree in debug and "release" mode and you'll see what I mean.

Peace,
pcfxer

[energyman]

FreeBSD had to compete with dead slow Ubuntu 8.10 (aka slow edition) and still lost.

What would happen if they had to compete with something fast?
It would be a massacre, that for sure.
And heavy IO? That is freebse's weakest points:

http://bulk.fefe.de/lk2006/bench.html

[kraftman]

I don't think so. The Linux kernel has currently a big problem with heavy io. There is an issue which in not really testable with the phoronix test suite. The desktop responsiveness is awful since 2.6.22. And it is not an ext3 issue. I have tried jfs and xfs.

Maybe your hardware is a problem, because I don't notice any responsiveness issue related to Linux kernel. Do you believe that such regression wouldn't be noticed for years?

@pcfxer

Fanboys comming... Are you dumb or something? Benchmark shows that Linux beat your FreeBSD... Idiotic attempt to make a flame war. If you're interested I can show you a benchmark in which Linux kernel 2.6.22 and newer one kick your lovely FreeBSD 7 or 7.1.xx.

And you registered to write such a bullshit? xd

EDIT:

@Bugmenot

I missed your post before. You're completely right!