Submitted by “Since the conclusion of the SMPng project, the focus of SMP development in FreeBSD has shifted from deploying locking infrastructure to careful profiling and optimization of kernel SMP strategies for increased performance on common workloads. FreeBSD 7.0 was the first release to benefit from this optimization work.” The status of this work includes MySQL workload benchmarks and memory allocator performance in the new FreeBSD 8 branch. Also, here is a recent presentation showing FreeBSD compared to several other operating systems like NetBSD, DrangonFly, Solaris, and Linux.
Pity that they did not compared that (great) performance to OpenSolaris or even Solaris SXDE, I am curious how Sun’s baby will perform here.
Like that one for example:
http://netbsd.org/~ober
I hope to test Solaris in the near future, the issue has been that I need to get it to boot via NFS on the test hardware.
It looks like glibc memory allocator needs some love.
—
http://www.kernel.org/pub/linux/kernel/people/npiggin/sysbench/
What, exactly, on that page leads you to believe the glibc memory allocator needs some love?
I’m not seeing it.
[edit] Doh, the URL is in your sig.
Edited 2008-03-11 15:05 UTC
Perhaps a bit of off topic but does anyone know if FreeBSD 7.0 supports the JMicron IDE controller?
Well, I got P35-DS3 mobo with that controller and it works fine, HDD that is. Didn’t try the DVD.
Edit: Added clarification to the usage.
Edited 2008-03-11 11:51 UTC
“Perhaps a bit of off topic but does anyone know if FreeBSD 7.0 supports the JMicron IDE controller?”
I have Asus P5B-VM with JMicron controller and it does problems on FreeBSD 7.0 with PATA DVD (burning/reading), PATA HDD seems to work fine.
Edited 2008-03-11 11:51 UTC
workaround solution! change IDE CDROM DMA mode to PIO mode.
In the boot prompt, type
hw.ata.atapi_dma=”0″
and add this in /boot/loader.conf
Thanks for working sollution, but I have already migrated to SATA, both HDD and DVD and disabled JMicron in BIOS.
This step was also mandatory to run/install Mac OS X or even OpenSolaris on this box.
They certainly seem intent upon proving that FreeBSD is faster than Linux. Can’t say as I blame them given how poorly FreeBSD has performed in the past. But the problem with trying to do that with benchmarks is that other people can run benchmarks right back at you:
http://www.kernel.org/pub/linux/kernel/people/npiggin/sysbench/
Notice how Nick isn’t claiming he has disproven the benchmarks, so you can’t claim that either
Actually he is using both different hardware (which may be relevant because this benchmark is very sensitive to CPU differences and Linux may be getting benefit from certain scheduler optimizations that are not yet enabled by default on FreeBSD) and different software (hand-compiled kernel with different configuration, and hand-compiled glibc). Most importantly he has not been able to replicate our results with older kernels, so we can’t conclude that there was a change in Linux that improved performance.
There is clearly something different about his test setup that needs to be understood.
I’ve gone to the FreeBSD.org; it would be nice if the updated their website. Their SMP page hasn’t been updated since 02/25/2007 – I”m sure something has happened in 1 year!
I remember a while back FreeBSD core developers were complaining that they couldn’t get enough programmers – and yet, when you look at their website, with the lack of regular updating, it screams, “there is no drive or passion in the project”.
I know I’m going to get slammed for this, but lets face the reality of the situation, if a budding programmer is looking for a project to contribute to – are they going to contribute to a project that appears (through the lack of updating the website) on deaths door or look at a project where there is constant buzz, communication and drive to inform the public where the project is heading.
Most real programmers do not look at websites, it’s something for the enduser at the desktop.
Even so; if they’re not going to update it, why have it? A ‘real programmer’ is that the same thing as a ‘real Scotsman Fallacy’? come on, if you’re a programmer or an end user, the first point of contact is the website, if no one can be bothered communicating information to the public via their front, then why even take an interest in either using the product or contributing to it.
Sure it’s to some degree nice for some people, but in the end it’s superficiality.
Most professional work is done on the mailinglists.
Before there was web, there was email and there were man pages. Some of us completely forgot to jump the wagon.
Anyway, if you’re offering a helping hand then welcome!
The web page says:
—
Project Status
[…] This web page is largely historical, and is not up-to-date with respect to the completion status of items in the task list, etc.
—
As for why this has not been updated: those of us who work on SMP are putting time into improving the code rather than the documentation. We’d be happy to work with someone who wants to improve the documentation, so please get in touch if you’re interested.
This is not true. There is enough drive and passion in the project. A quick look at what was accomplished with 7.0-RELEASE is proof enough.
The website: It is rebuilt daily. Obviously, not all sections are updated very regularly. Keep in mind that a lot of the more technical stuff needs to be written by the people who understand them better: the coders. They are probably just hacking away now
Documentation of FreeBSD tends to be of the best kind around, compared to any OS, commercial or not. Have a look at the articles, the handbook etc. These are in constant development and a lot of people are helping. Follow the mailing lists like -questions and -doc for a while and you will see what I mean.
“I’ve gone to the FreeBSD.org; it would be nice if the updated their website. Their SMP page hasn’t been updated since 02/25/2007 – I”m sure something has happened in 1 year!”
On the contrary, I find the FreeBSD website one of the more commonly updated websites. What you are referring to is a legacy page that has been kept there for historical reasons only. Since 2/25/2007, the SMP project has been deemed successfully completed and that page will no longer be updated. If you want a status of what has happened and what needs work, go here: http://wiki.freebsd.org/SMPTODO
“I know I’m going to get slammed for this, but lets face the reality of the situation, if a budding programmer is looking for a project to contribute to – are they going to contribute to a project that appears (through the lack of updating the website) on deaths door or look at a project where there is constant buzz, communication and drive to inform the public where the project is heading.”
What project the programmer chooses really depends on what kind of programmer the budding programmer wants to be. If he wants to be a real programmer, the choices are pretty limited to kernels, compilers, languages, and databases. If you want to do kernels, I personally think FreeBSD stacks up better against Solaris and Linux as it is also an advanced kernel that gives you the opportunity to make an impact within the project in terms of both code and design, that allows you to work in a true team/project setting, and that provides for a consistent vision throughout. Unless you work for Sun, you will be basically restricted to bug fixes for life. Similarly, Linux development is such a mess with poor standards and a kernel that changes rapidly without reason, that also will lead any new developer to a life of bug fixes.
>if a budding programmer is looking for a project to contribute to
He will usually knock on the door of a suitable mailinglist as usual in every single free operating system, even in Solaris community.
With the help of Nokia,
http://jeffr-tech.livejournal.com/19139.html
“Lately I’ve been able to spend a bunch of time on ULE thanks to Nokia. They use it in one of their networking products. I’ve been doing all of this work in 8.0-CURRENT and backporting it for them at the same time. It’s a great model for both parties because users on -CURRENT shake out bugs that they’d have to find in testing otherwise and we get new development paid for.”
The only thing I found surprising in the benchmarks was WRT NetBSD. My own damned fault for not having paid attention to its development over the last year or so.
I would expect that Linux and FreeBSD will be playing cat and mouse scalability wise for the foreseeable future.
DragonFly’s poor scalability in the tests is _not_ unexpected as it still runs under the giant kernel lock, and from benchmarks done by Kris on DragonFly 1.12 (see the DF Users mailing list archive), things aren’t much different.
> I would expect that Linux and FreeBSD will be playing cat and mouse scalability wise for the foreseeable future.
I hope so – this spells improvements all around
I would expect that Linux and FreeBSD will be playing cat and mouse scalability wise for the foreseeable future.
I see this as a good thing.
What’s the point it this?
This is the third or fourth time we see those benchmarks. Ok, ebizzy and bind are new, but what’s the point? The ISC did their own bind benchmarks, take a look at http://new.isc.org/proj/dnsperf/OStest.html
Ok, they are using older versions.. fair enough, but at least they are independent.
The ebizzy benchmark indicates that glibc’s memory allocator isn’t the best
As for MySQL benchmarks: http://www.kernel.org/pub/linux/kernel/people/npiggin/sysbench/
Linux actually performs better..
You had your 5 minutes of fame, the SMPng project took 5 years longer than you expected and now you’re trying to show everyone that you are faster than Linux.. please..
Really what’s the point? Of course you pick benchmarks where you might perform better, I can’t blame you for that, but when was the last time some Linux kernel developer came out and posted a benchmark where FreeBSD didn’t do so well? You have to admit that they had plenty of time between 4.x and 7.x ;D
Now something positive.. the first MySQL benchmark published last year was actually good, because you exposed a real bug in glibc, but after that was fixed it’s all a bit picky..
I think the point is to say “hey guys we’re back in the game”.
“Nick Piggin has been doing some benchmarking of recent linux kernels and FreeBSD 7.0 on a 2xquad core barcelona opteron. He verified that the CFS problems seem to be fixed and FreeBSD’s performance on this box with mysql is really very similar up to about 20 threads. I feel confident that the test was conducted fairly and I’m happy with these results. Our stable release is doing very well even if fresh-out-of-git linux is showing better on this platform.” <– Jeff R. on the recent linux is faster benchmark.
It’s just friendly competition — and the only people who are turning this into dick size contests are those fanatics who are reading a little too deeply into it.
Everyone knows benchmarks are only relevant within context. The FreeBSD guys are comparing to linux because linux has been considered a performance leader the last few years… it’s the easy first choice. It wouldn’t make much sense for them to compare to OpenBSD in terms of performance and scalability, now would it?
>Linux actually performs better..
You should take 1 of the five minutes and actually read your nice *benchmark*. There is no war, just some FUD in the head of some people.
>but when was the last time some Linux kernel developer came out and posted a benchmark where FreeBSD didn’t do so well?
Oh I forgot something? There is at least always a developer of Linux who spreads FUD about other operating systems. But in the end I wouldn’t post a benchmark at all being a Linux developer, because we could actually see the real ‘performance’.
There:
http://www.linux.com/articles/45571
or there too:
‘I claim that Mach people (and apparently FreeBSD) are incompetent idiots. ‘
http://www.uwsg.indiana.edu/hypermail/linux/kernel/0604.2/1219.html
So please stop the FUD if something like Linux is actually made by childish developers. You will _not_ find such an infancy in *BSD!
There:
http://www.linux.com/articles/45571
Oh well.. I don’t see him spreading FUD in there. He clearly says that he doesn’t know *BSD and that he thinks those discussions are pointless..
Some examples:
Torvalds: I really don’t much like the comparisons. In many ways they aren’t even valid, since “better” always ends up depending on “for what?” and “according to what criteria?”.
NF: If the BSDs were better technically five years ago, has the playing field leveled since then?
Torvalds: I don’t think they were better five years ago (see above), and I don’t think the question really makes sense.
NF: Are there parts of BSD today that you would like to see adopted in the kernel?
Torvalds: I certainly don’t have any specifics, but that’s not saying that I’d be against it. It just means that I don’t know anything about BSD technical internals, so I’m the wrong person to ask. Ask somebody who uses both.
I think Linus answers are pretty good actually..
‘I claim that Mach people (and apparently FreeBSD) are incompetent idiots. ‘
Well.. don’t put too much into it. He said apparently not certainly
Maybe he didn’t exactly FUD, still went ranting on how BSD is crap (the usual shtick: he doesn’t like it, ergo it’s shit), just to finally admit that he actually doesn’t have a clue about it.
Thanks, but I don’t want to like that guy or take him serious.
For people like you it is not worth it and it’s all about fame. But for caring linux devs it was something else. FreeBSD devs have been pointing out weakness left and right in the linux kernel performance and without them God knows when linux devs would have discovered these bottle necks! Consequently, hard working linux devs patched them and improved these points. So, move along.
For all the people who pointed Nick Npiggin website.
Here is a quote from his website
Now let’s see Nick’s Benchmarks on FreeBSD 8-CURRENT since he is benchmarking with an “unstable” linux rc.
[EDIT]
And for the record after reading a few posts. It seems to me (as always) that most of these Linux zealots are pretty childish. “Oh who said so, my car is faster than yours!”, “Oh you got your fame! Now look how I beat you.” Check out http://jeffr-tech.livejournal.com/18706.html Jeff’s friendly posts regarding the benchmarks! This is the FreeBSD community manners!
Not RTFM and sucking on lolly pops zealots.
Edited 2008-03-11 21:24 UTC
While you were quoting Nick, why did you not quote this:
“””
The Linux kernel used is not a “stable release” whereas FreeBSD is (although I’m not aware of any significant performance improvements over the 2.6.24 kernel — 2.6.25-rc4 is simply what I have installed on the machine)
“””
Or, we could just wait until next month when 2.6.25-final is released. Why should we test a FreeBSD kernel version which might not be ready for years? And, of course, Nick also included the rather old 2.6.22 kernel from last July, and it beat FreeBSD 7.0, as well.
But don’t you think the whole back and forth in this thread is kind of silly? I’m happy that both Linux and FreeBSD are doing well. And I’m glad we got a nice glibc bug fix out of all this.
My only annoyance has been the way the FreeBSD folks have paraded Kris’ benchmarks around ad nauseam for the past couple of months.
Edited 2008-03-11 21:42 UTC
I’m just happy with the performance improvements of FreeBSD, as a user of it. If Linux is faster or better at some workload, well, let’s use it as a good measure and try to achieve better performance.
This competition is GOOD for both projects. FreeBSD 5.x-6.x stablished a rock solid base for the future.
I do not need to quote it. Cuz It does not server any purpose in my argument nor it does mean anything. FreeBSD 8-CURRENT branch is active and can be used for benchmarks if you like. If it’s going to be a future release then once it is released compare it with its equivalent FreeBSD dev.
Regarding 2.6.22, you had a similar post in the previous thread. I replied to you here http://osnews.com/thread?303772 there is no need to repeat it again because there are several benchmarks on 2.6.22 which show different results.
No need to be annoyed
All FreeBSD devs have worked hard to bring FreeBSD back on top. In order to let the world hear it, such presentations and benchmarks are needed and justified. He has all the raw data as well as the hardware so If you think he is not telling the truth then you might want to ask the linux devs who worked with him to withdraw their patches from the linux kernel tree to see the effecth of his constructive and friendly criticism.
Edited 2008-03-12 02:51 UTC
What you keep missing is that the *difference* is actually in the glibc versions. You keep talking about the Linux *kernel* improving to match FreeBSD. When, in fact, the problem was in glibc-2.6. There is no inconsistency in the benchmarks. 2.6.22 beats FreeBSD 7.0 with glibc-2.7, and does not with glibc-2.6. I believe there was some kernel regression in 2.6.23 when Linux switched schedulers. It would be interesting to see 2.6.24 results. But I imagine we will see 2.6.25 final results next month, at which point your claim that we should be testing against the FreeBSD8 development version will be moot.
And no, no one in their right mind would run the FreeBSD 8 development version on a production server right now… or a month from now.
However, I’m very happy to see that FreeBSD is back up to snuff after all this time. I’m sure that a lot of people worked very hard on that come-back.
FYI, I have been trying hard but cannot replicate the ISC benchmarks, even using identical data sets and test methodology. My best guess is that since their test hardware is using the bge driver — which is a mass of quirks and workarounds for broken broadcom hardware in both linux and FreeBSD — then perhaps FreeBSD is missing bug workarounds for their particular model.
http://people.freebsd.org/~kris/scaling/bind-pt.png
http://people.freebsd.org/~kris/scaling/bind-pt-gige.png
I see FreeBSD outperforming Linux by about 35-40% with Intel gigabit ethernet and chelsio 10 gigabit ethernet drivers. The performance drop above 6 named threads is a scalability problem in named.
As for why we’re doing these comparisons: if you don’t know where you perform well compared to other OSes, you don’t know where you perform badly either, and that tells you where you should concentrate future work. This works in both directions, and Linux is also benefiting from the comparisons. The end result is a win for the users.
Linux may not know about BSD, but he is, occasionally at least, funny.
Any news on how good this ‘experimental’ support is?
I’m hoping to run a home server runing FreeBSD and ZFS – i just want to know that my data will be safe (ignoring, for the moment, the obvious need to back up data)