FreeBSD Project Launches FreeBSD 6.0 [LWN.net]

FreeBSD Project Launches FreeBSD 6.0

Posted Nov 5, 2005 0:16 UTC (Sat) by tomsi (subscriber, #2306) [Link]

Ooops. Was a bit quick there.

What I tried to say, was:

The most interesting part of the "blurb":

"Recent performance benchmarks show that
FreeBSD 6.0 outperforms Linux in raw data throughput."

Tom

FreeBSD Project Launches FreeBSD 6.0

Posted Nov 5, 2005 2:08 UTC (Sat) by Los__D (subscriber, #15263) [Link]

Quite a statement without links to the benchmarks to back up their claim, though...

FreeBSD Project Launches FreeBSD 6.0

Posted Nov 5, 2005 4:46 UTC (Sat) by cventers (subscriber, #31465) [Link]

Yeah, I'm all in favor for competition, but I remember FreeBSD pretty
much becoming irrelevant around 2.4 Linux, with 2.6 Linux advancing so
much as to wipe the floor with all of the BSD's...

I'll believe it when I see the numbers.

FreeBSD Project Launches FreeBSD 6.0

Posted Nov 5, 2005 13:56 UTC (Sat) by kervel (guest, #9133) [Link]

i'm having a wireless card for over a year now, and still i need to patch my kernel, another
computer with an atheros wireless card has to be patched too. Seems the BSD's are definately
better for this.

FreeBSD Project Launches FreeBSD 6.0

Posted Nov 5, 2005 14:34 UTC (Sat) by arcticwolf (guest, #8341) [Link]

While it certainly is good that (Free)BSD seems to have solid support for wireless cards, that doesn't really have anything to do with superior disk performance, though.

FreeBSD Project Launches FreeBSD 6.0

Posted Nov 5, 2005 22:41 UTC (Sat) by b7j0c (subscriber, #27559) [Link]

so you are requesting that we accept your baseless suppositions instead of theirs? "no test data" does not mean "linux wins".

FreeBSD Project Launches FreeBSD 6.0

Posted Nov 6, 2005 2:38 UTC (Sun) by cventers (subscriber, #31465) [Link]

Take a look at the other benchmarks linked on this page. Also, find an
old NewsForge article comparing the some BSDs to Linux to Solaris in SMP
MySQL performance.

(Hint: If there were no Linux, I'd happily choose Solaris over BSD).

FreeBSD Project Launches FreeBSD 6.0

Posted Nov 6, 2005 3:58 UTC (Sun) by b7j0c (subscriber, #27559) [Link]

>> Take a look at the other benchmarks linked on this page.

yeah, from someone else's post, so don't claim them as your "source" after the fact.

in any case that survey compares the 2.6 kernel and freeBSD 5.1 quite favorably. it would be useful if this were more up to date, but it is very detailed.

FreeBSD Project Launches FreeBSD 6.0

Posted Nov 6, 2005 7:13 UTC (Sun) by cventers (subscriber, #31465) [Link]

I was familiar with the 'fefe' benchmark and have been for some months
(remember, it's old enough to have tested Linux 2.6.0-test7). If you're
going to disagree with me, do it on factual grounds...

>> yeah, from someone else's post, so don't claim them as your "source"
after the fact.

I'm sure you don't believe that it is possible I've actually seen that
benchmark before. Yet, regardless, this statement of yours is totally
worthless as it is a direct attack on me versus sane discussion of the
subject matter.

FreeBSD did seem to do fairly well for most of the benchmarks, indeed.
However, look at the Newsforge MySQL benchmarks - all three BSDs suffered
terribly in the jump from 1 to 2 CPUs.

As a side note, NetBSD turned out to be misconfigured, because at the
time there was an *undocumented* gotcha that you had to set
PTHREAD_CONCURRENCY in the environment in order to actually allow the
software to go parallel. Yet this did not save the day for the other two
BSD's.

The bottom line on this performance argument is this:

1. Linux beats FreeBSD in nearly every benchmark I've seen for a while,
either by a little or by a lot;
2. The above is especially true when the number of CPUs starts to rise.

FreeBSD Project Launches FreeBSD 6.0

Posted Jan 10, 2006 0:41 UTC (Tue) by jtanis (guest, #35077) [Link]

Your proving your own ignorance here. 5.x series didn't even become stable until 5.4. Much of the work that was done, was on removing the GIL to improve SMP (as well as UP performance) down the road. The early 5.x versions were never really expected to win any performance tests, they were a *transitionary* period. At this point in time (which was like.. what, 3/4 years ago?) the 4.x series still outperformed the 5.x series. Fact of the matter is, if your "proof" is older than an operating systems last two stable releases (by a number of years) your not using accurate information.

NetBSD

Posted Nov 6, 2005 9:50 UTC (Sun) by hummassa (subscriber, #307) [Link]

was faster in raw data performance than Linux 2.6

FreeBSD Project Launches FreeBSD 6.0

Posted Nov 10, 2005 9:39 UTC (Thu) by ekj (subscriber, #1524) [Link]

It's actually not that hard.

For example, a multi-threaded filesystem generally leads to lower throughput.

Logical enough. Fulfilling multiple IO-requests in parallell requires switching between those requests, which takes time. Fulfilling one request first, then going on to the next would give better throughput.

Offcourse it would also mean that you "ls" never got the 1KB directory-read it needs before the 1GB "cp -a somedir /somewhere" process finished.

Thus, "has higher troughput" can well be true (I don't know if it is) but even if it *is* true it doesn't, for most workloads translate into "is better".

FreeBSD Project Launches FreeBSD 6.0

Posted Jan 10, 2006 0:29 UTC (Tue) by jtanis (guest, #35077) [Link]

Bah, what you remember is skewed with actual reality. Their wasn't even a moment that Linux even came close to making FreeBSD irrelevant much less "wiping the floor with it." Heck, I actually had better SATA support from the FreeBSD 5.x series then I ever got from Linux's 2.4 kernel. I'm primarily a FreeBSD user, but I use both OS's depending on various factors. Every once in a while I run my own benchmarks, last time I did this I typically saw better memory throughput with FreeBSD, but better CPU performance with Linux. If you do a little research on implementation, this isn't actually too unexpected. Your mileage may vary, but we don't have to prove it to you. It's my opinion that close minded generalizations such as yours don't belong in the open source community.

FreeBSD Project Launches FreeBSD 6.0

Posted Nov 5, 2005 5:46 UTC (Sat) by nurhussein (guest, #16226) [Link]

That also struck me, I've *never* seen any benchmark numbers that support the oft-repeated claim of BSD's alleged superiority. I'd like to see some, preferably in a peer-reviewed publication.

FreeBSD / Linux benchmarks?

Posted Nov 5, 2005 15:05 UTC (Sat) by shane (subscriber, #3335) [Link]

I agree. There aren't too many studies that Google finds. There was a really nice one a few years ago:

http://bulk.fefe.de/scalability/

Linux kicked ass here.

You will also hear FreeBSD fans mention vague things about FreeBSD having a "better network stack" and such. Yet, if you look at things like the Internet2 Land Speed Record:

http://lsr.internet2.edu/history.html

You see a lot of Linux boxes, some Windows boxes, and even NetBSD. But no FreeBSD.

The Volcano Report is a few years old, and punishes FreeBSD because FreeBSD historically has had a pretty lame thread model, but here we also see Linux and Windows outclassing FreeBSD:

http://www.volano.com/report/

I'm not saying that FreeBSD isn't fast, I am just saying I have never seen a single quantitative sutdy that backs that up. And in my experience, FreeBSD performs poorly, although I freely admit my FreeBSD experience is limited - I gave up on it after several frustrating failures.

FreeBSD / Linux benchmarks?

Posted Nov 5, 2005 15:36 UTC (Sat) by samb (subscriber, #32949) [Link]

FreeBSD advocacy is all about parroting memes. The "better network stack" claim is a relic of the early 90s, and basically hasn't been true for ten years.

Performance v.s. Quality

Posted Nov 5, 2005 17:27 UTC (Sat) by mheily (guest, #27123) [Link]

Both systems have similar performance characteristics. If you actually read the http://bulk.fefe.de/scalability/ study, FreeBSD and Linux both scale O(1) in most of the tests. The study concludes that:

"FreeBSD 5.1 has very impressive performance and scalability... FreeBSD has by far the best performance of the BSDs and it comes close to Linux 2.6."

Instead of focusing solely on performance, Linux afficionados should recognize the importance of other factors such as good documentation, stability of programming interfaces, backwards compatibility, designing for the future, source control, and release engineering. FreeBSD produces a high quality UNIX operating system by following these principles.

Take a look at FreeBSD subsystems like GEOM, DevFS, kqueue, and then look at the history of their counterparts in Linux (devfs/udev, dnotify/inotify). It seems like FreeBSD gets it right the first time, while it takes Linux several attempts to find the right solution.

Both systems can coexist and learn from each other, and I'm glad to see that FreeBSD has continued to improve despite being overshadowed by the growth of Linux.

Performance v.s. Quality

Posted Nov 5, 2005 18:09 UTC (Sat) by khim (subscriber, #9252) [Link]

FreeBSD produces a high quality UNIX operating system by following these principles.

Heh. May be. But in the end Linus said it the best:

But exactly because Linux tries to be "good enough" for everybody, you'll find a lot of areas where Linux is better (often a lot better -- as in "it works"), and then you'll find a few narrow areas where one particular BSD version will be better.

If it does not work then all other advantages are useless - and unfortunatelly this happens way to often with freebsd. At least in my own expirience. Looks like FreeBSD 6.0 is step in right direction but only time will tell - is it real or is it just hype.

Performance v.s. Quality

Posted Nov 5, 2005 21:19 UTC (Sat) by cventers (subscriber, #31465) [Link]

I'm not convinced. While I don't believe that the 'quality' of a product
necessarily reflects its adoption directly, I think there is a big reason
Linux has gotten as huge as it has, leaving FreeBSD in the dust, so to
speak. Why is it taking FreeBSD so long to benefit from SMP?

True, Linux has always been more of an experimental system, but I don't
think anyone can say it's been bad for the project.

You want to know why I think we're in the state we are today? Because
FreeBSD held on to all of the rules, and Linus and his band of hackers
threw them right out the window.

Until FreeBSD can start offering consistent and practical advantages over
Linux (as in, not 'this is better in theory' or 'this is better on
principal') then I'm just not going to be convinced.

Performance v.s. Quality

Posted Nov 5, 2005 22:46 UTC (Sat) by b7j0c (subscriber, #27559) [Link]

your post is 100% content-free zealotry.

compare "make world" to updating facilities in any distro with perhaps maybe a nod to debian (which gets trashed for being too old-fashioned).

Performance v.s. Quality

Posted Nov 6, 2005 2:50 UTC (Sun) by cventers (subscriber, #31465) [Link]

Have you ever used Gentoo?

Quite frankly, I see you doing a lot of dancing around in these comments.
Just a few posts down you're making excuses for the rotten state of
FreeBSD 5.

I'd like to repeat the crux of what I said:

Until FreeBSD can start offering consistent and practical advantages over
Linux (as in, not 'this is better in theory' or 'this is better on
principal') then I'm just not going to be convinced.

Can you point these consistent and practical advantages out? So far,
Linux holds tons of seats on the Top 500 supercomputing sites in the
world; offers tons of distributions to choose from with *no fork* in the
kernel or core OS (compare that to having to choose between the merits of
Free/Open/Net/DragonflyBSD... they often claim that these BSDs represent
differences in philosophy, when in cases like OpenBSD it was the result
of egos and assholes).

So what am I choosing when I choose between Linux distributions? If I
want an out-of-the-box supported no-frills enterprise Linux, I might
choose Red Hat. If I want a stable server OS and can deal with guru
management, I might choose Slackware or Debian. If I want a total
tinkerbox, I choose Gentoo.

Regardless of whether I choose Red Hat, Slackware, Debian or Gentoo, I'm
running some flavor of 2.6 Linux. 2.6 Linux provides me with outstanding
performance in nearly all corners of the OS (indeed, 2.6.0-test7
demonstrating O(1) scalability in every benchmark conducted in one of the
links you'll find in this article's comments), tons of filesystem
flexibility in terms of LVM, md, and a slew of supported filesystems
(indeed, I could choose JFS, ext2/3, XFS, reiserfs for a root
filesystem), excellent hardware support (much better than any other open
source OS I am aware of), native SMP support scaling up to *huge*
systems, an O(1) process scheduler, etc, etc, etc, etc, etc.

I don't really mean to come in here and act like a troll. I'm just sick
and tired of hearing FreeBSD zealots repeating tired old claims of
superiority from years and years ago that are simply no longer true. None
of your ranks, in any conversation in which I have participated, have
ever concretely demonstrated or began to demonstrate any validity in
their arguments.

Instead, it seems that the noise I receive from the BSD community
includes things like Theo making arbitrary blanket statements in the
press like "Most people don't know how terrible Linux is"... (not that
Theo speaks for the BSD community at large).

Bottom line? (Addressed to the FreeBSD project) Stop trying to assert in
public that you're better than Linux until you can prove it.

Performance v.s. Quality

Posted Nov 6, 2005 4:18 UTC (Sun) by b7j0c (subscriber, #27559) [Link]

>> Can you point these consistent and practical advantages out?

once again, better man pages, superior code delivery solution in /ports, and in my opinion, much better performance given heavy loads.

i am not knocking linux. i also use Ubuntu linux, which is the first distro in my opinion to approach freeBSD's "system" approach.

>> Linux holds tons of seats on the Top 500 supercomputing sites

haha are you serious? these installations are so tweaked they bear little resemblance to your Gentoo box.

these people use linux because of beowulf, not because it is "better" than freeBSD.

>> offers tons of distributions to choose from with *no fork* in the
>> kernel or core OS

freeBSD is not "forked" either. you don't see some people running "freeBSD with featureX" and "freeBSD with featureY". dragonfly is a completely separate codebase at this point, and there is no suggestion that these projects support each other. the other BSDs are entirely different projects.

now as for "forking" or slight variations in "what you are running", let me assure you the linux distribution landscape is ALL OVER THE MAP. some use different /rc layouts. some only support installing from source. blah blah blah. this is exactly why the LSB exists, because the linux world is in far greater danger of manifesting incompatibilities than the freeBSD world.

this is because freeBSD is not a kernel, it is an operating system.

>> when in cases like OpenBSD it was the result of egos and assholes

you're purely amateur hour at this point. please head on over to the openbsd boards and let them know that their detailed security work is absolutely worthless (you might want to read up on rudimentary OS security first), its all just ego.

>> Regardless of whether I choose Red Hat, Slackware, Debian or Gentoo, I'm
>> running some flavor of 2.6 Linux

this hasn't always been the case. debian was 2.4 for quite a while. i'm not saying this was a bad decision on their part, simply pointing out that the linux world is rife with compatibility issues and minor nuances that do not exist to the same extent on freeBSD.

>> tons of filesystem
>> flexibility in terms of LVM, md, and a slew of supported filesystems
>> (indeed, I could choose JFS, ext2/3, XFS, reiserfs for a root
>> filesystem), excellent hardware support (much better than any other open
>> source OS I am aware of), native SMP support scaling up to *huge*
>> systems, an O(1) process scheduler, etc, etc, etc, etc, etc.

you talk as if freeBSD does not have a good filesystem, good process scheduling etc.

>> Bottom line? (Addressed to the FreeBSD project) Stop trying to assert in
>> public that you're better than Linux until you can prove it.

show me where i said "freeBSD is better than linux"

what i said was that it has some features that are superior to what exists on many linux OSs (make world, ports, man pages etc)

i also said it is very solid as a server OS

i also said i use Ubuntu linux

i think you should actually use freeBSD before you trash it, and in general get a clue before coming in here on lwn and telling people like openBSD devels that they are assholes (have you ever even met these people???).

Performance v.s. Quality

Posted Nov 6, 2005 5:45 UTC (Sun) by tzafrir (subscriber, #11501) [Link]

b7j0c wrote:

> > Can you point these consistent and practical
> > advantages out?

> once again, better man pages, superior code delivery
> solution in /ports,
[snip performance claim]

Please check ports as well. Unlike Debian that has a strict policy for not recieving programs with no man pages (e.g: who wrote the man pages for KDE?), FreeBSD has no problem with those.

> i am not knocking linux. i also use Ubuntu linux,
> which is the first distro in my opinion to approach
> freeBSD's "system" approach.

You managed to miss Debian.

Performance v.s. Quality

Posted Nov 7, 2005 6:17 UTC (Mon) by mepr (guest, #4819) [Link]

being a debian user, I'll be the iconoclast and point out that some of the debian man pages consist of something equivelant to the following:

"um, yeah.
We haven't gotten around to writing this.
It's fine if you write one for us.
Thanks"

Performance v.s. Quality

Posted Nov 7, 2005 10:22 UTC (Mon) by syntaxis (guest, #18897) [Link]

Which ones? Bugs should be filed against those packages...

Performance v.s. Quality

Posted Nov 6, 2005 7:32 UTC (Sun) by cventers (subscriber, #31465) [Link]

>> once again, better man pages, superior code delivery solution
in /ports, and in my opinion, much better performance given heavy loads.

Having not spent a lot of time in the manual pages, I'm not going to
argue that point one way or another. Superior code delivery system I'm
going to have to disagree with, given that I'm now using Gentoo's portage
on every system I control. As for "much better performance given heavy
loads", perhaps you can point to recent benchmarks that demonstrate this?
I spent some time in freebsd-performance and I seem to recall lots of
posts arguing about things like, "Why does MySQL perform *so much better*
on Linux?" (so much better being 2x+)

>> haha are you serious? these installations are so tweaked they bear
little resemblance to your Gentoo box.

Some installations may be specialized, but that fact as a blanket
statement is simply pulled out of your ass. This article, linked below,
talks about HP's experiences running stock Linux on a 64-node Itanium2
superdome:

http://news.zdnet.co.uk/0,39020330,39184546,00.h

I quote:

>> Running three different benchmarks on a standard Linux distribution
based on the 2.6 kernel, the Superdome showed linear improvements for
kernel compiling, memory bandwidth, and the HPL common supercomputer
benchmark.

Try that with FreeBSD and report back.

>> The HPL benchmark, which is used to measure performance when solving
large linear equations, produced similar results, rising from 18
gigaflops with one cell of four processors to 277 gigaflops with all 16
cells, or 64 processors, running.

>> "This was a standard Linux distribution," said Cabaniols. "The kernel
was able to discover the topology of the system and discover the memory
in a NUMA pattern."

>> freeBSD is not "forked" either. you don't see some people running
"freeBSD with featureX" and "freeBSD with featureY".

No, FreeBSD is not internally forked. If it were, they might have an
easier time porting features. Instead, you have NetBSD, FreeBSD, OpenBSD
and DragonflyBSD which share ancestry but can't directly share code
because they are all *different kernels*. In Linux, you have lots of
distributions to choose, but all of them ship kernels that are directly
based on Linus's tree.

>> you're purely amateur hour at this point. please head on over to the
openbsd boards and let them know that their detailed security work is
absolutely worthless (you might want to read up on rudimentary OS
security first), its all just ego.

Are you going to try to tell me Theo is not an asshole? Because you might
want to research why OpenBSD exists in the first place. Theo's CVS access
from NetBSD was forcibly removed:

(quoting Wikipedia)
In December 1994, Theo de Raadt, a co-founder and member of the NetBSD
core team for two years, was asked to resign from the NetBSD Foundation.
His access to the NetBSD CVS server was terminated and he was instructed
to e-mail any further changes to the system as patches, so that the core
team could check them. He was also informed that he no longer represented
the NetBSD project in any formal manner.

I'm not saying that BSDs don't contribute anything. I think OpenSSH, for
example, is an excellent project. What I was stating is that the reason
OpenBSD exists is not because Theo woke up one day and decided to create
a different distribution... it's because his personality was abrasive
enough for the rest of the NetBSD team to toss him to sea.

>> this hasn't always been the case. debian was 2.4 for quite a while.
i'm not saying this was a bad decision on their part, simply pointing out
that the linux world is rife with compatibility issues and minor nuances
that do not exist to the same extent on freeBSD.

Yes, and there was a time that FreeBSD 4 didn't exist! What the hell is
your point, exactly? Because "debian was 2.4 for quite a while" doesn't
say anything at all about compatibility issues or minor nuances - you're
totally talking out of your ass here.

>> you talk as if freeBSD does not have a good filesystem, good process
scheduling etc.

I'll call this pure opinion to avoid having to dig up literature on the
subject; but I do believe FreeBSD has a "good" filesystem and a "decent"
process scheduler. I also think that Linux has "lots of excellent"
filesystems and an "excellent" scheduler (and indeed, with patches on
their way in, "7 excellent schedulers"). Nevermind that Linux lets you
choose I/O schedulers on the fly on per-block-device granularity.

A few final points:

1. You assume I've never used FreeBSD. This is an assumption, because in
fact, I have used FreeBSD and have also assisted UNIX-novice friends with
FreeBSD for web server purposes.

2. The whole reason I've replied to this article is in response to the
claim in the FreeBSD press release that "a recent benchmark shows FreeBSD
having faster data throughput". Not saying there isn't a way they could
actually pull ahead in the race once in a while, but given the amount of
blind FreeBSD zealotry that is evident in the community (see: picture of
Tux getting raped by BSD daemon), and the number of times people make
unsubstantiated claims about FreeBSD being faster/better, I'm not
inclined to believe *any* claims of FreeBSD being faster at anything
until someone can demonstrate it in an objective benchmark.

And really, it doesn't matter. I enjoy Linux when I'm in Max Payne 2 or
Doom 3 or Quake 4 or UT2004 or Starcraft, or when I'm surfing the
internet with Konqueror, working with KDE, coding/doing web development,
or watching my cluster flawlessly handle lots of e-mail / pageviews /
database queries / telephone calls. I'm happy to coexist in the community
with FreeBSD (indeed, it's always good to have good competition).

Bottom line? I'm sick and tired of people blindly touting FreeBSD's
superiority without any way to demonstrate any actual basis in fact. Man
pages and release engineering don't help you crunch more numbers or serve
more web pages, and in the server capacity, that's what this is all
about.

Performance v.s. Quality

Posted Nov 6, 2005 10:13 UTC (Sun) by cventers (subscriber, #31465) [Link]

Apologies in advance... the ZDNet link needs to end in "htm", not "h"

Performance v.s. Quality

Posted Nov 7, 2005 4:35 UTC (Mon) by b7j0c (subscriber, #27559) [Link]

>> you have NetBSD, FreeBSD, OpenBSD
>> and DragonflyBSD which share ancestry but can't directly share code
>> because they are all *different kernels*

they are more than different kernels, they are different operating systems and different projects.

they have made no claims of portability between OSs, this is your hang up.

i have no idea why you are hanging this claim on them...why, because they all started with 4.4bsd back in what, 1991?

on that note i would like to declare severe problems with linux and solaris compatibility, after all, they are both written in C and support processes and work in intel systems.

>> Are you going to try to tell me Theo is not an asshole?

i don't know him. also if i want to say something like that to someone, i do it directly to them and with good reason, which is why i suggested you do so, signed with your real name of course (since you are insulting him by name). presuming that all you know about him is what you pulled out of wikipedia, you have no cause for insulting him in a public forum either.

>> Bottom line? I'm sick and tired of people blindly touting FreeBSD's
>> superiority

you keep syaing that, even though no one has.

Performance v.s. Quality

Posted Nov 7, 2005 4:50 UTC (Mon) by cventers (subscriber, #31465) [Link]

>> they are more than different kernels, they are different operating
systems and different projects.

Indeed, you're right! I'm declaring this to be a *problem* with BSD,
because innovation cannot be easily shared. What if all the developers
for Linux split the kernel into three separate projects? I'd speculate
that each project would still have more manpower than the most developed
BSD, but it would do seriously bad things to Linux's rate of innovation,
which I believe the forking has done to BSD.

>> i don't know him. also if i want to say something like that to
someone, i do it directly to them and with good reason, which is why i
suggested you do so, signed with your real name of course (since you are
insulting him by name). presuming that all you know about him is what you
pulled out of wikipedia, you have no cause for insulting him in a public
forum either.

You pick some really clueless vectors on which to attack me. What makes
you think that my first exposure to Theo is on Wikipedia? Don't you think
there was a reason I thought that might be a good place to look for
someone else's words on the subject?

Indeed, if you had any kind of short term memory, you might remember when
I said:

>> Instead, it seems that the noise I receive from the BSD community
includes things like Theo making arbitrary blanket statements in the
press like "Most people don't know how terrible Linux is"... (not that
Theo speaks for the BSD community at large).

Read this article:

http://www.forbes.com/intelligentinfrastructure/2005/06/1...

I, Chase Venters, hereby declare Theo De Raadt to be an asshole on
account of his childish slander that seems to carry no more substance
than sour grapes.

>> you keep syaing that, even though no one has.

I do keep "syaing" that, because of the claim that sparked this whole
flash of commentary - the statement in FreeBSD's press release that some
unspecified benchmark demonstrates that FreeBSD 6 is faster than Linux in
data throughput. I don't generally believe unqualified numbers, and I
certainly don't believe unqalified numbers from the FreeBSD project.

I'm going to voluntarily stop adding comments to this thread, because I'm
starting to feel like a Slashdot troll, and I'm quite fond of LWN as an
excellent resource.

Performance v.s. Quality

Posted Nov 7, 2005 5:49 UTC (Mon) by b7j0c (subscriber, #27559) [Link]

>> Indeed, you're right! I'm declaring this to be a *problem* with BSD,
>> because innovation cannot be easily shared.

each project has separate design goals. they are happy to sail in separate directions. these design goals are oftne mutually exclusive. openbsd's performance may not be optimized, but its security features are. i would offer that openbsd is generally considered the most secure open source OS, so forsaking comprimise has worked in their favor. likewise with netbsd's portability. they don't want to adopt code that will forsake this design goal.

the entire point of the BSDs is to provide an implementation of a specific design goal, not to conflate all design goals.

Performance v.s. Quality

Posted Nov 7, 2005 10:49 UTC (Mon) by PaXTeam (subscriber, #24616) [Link]

as someone else mentioned it already, this is just another meme that some (Open)BSD users keep repeating. what is it exactly that allows a Linux or (say) NetBSD system to be compromised but doesn't allow it on OpenBSD? can you mention a few specific things or have you blindly repeated their propaganda only? and i'd also like to know who 'in general' considers OpenBSD the 'most secure open source OS'? i've spent quite some years in computer security and haven't run across any expert who thinks so. i submit we may differ in the definition of 'expert' here, to me it's the person who knows how to avoid/find/exploit/fix bugs, and has actually done so in a professional manner for a number of years, i.e., actual real-life experience is king and key here (and it excludes quite a few self-described 'experts').

Performance v.s. Quality

Posted Nov 7, 2005 17:01 UTC (Mon) by b7j0c (subscriber, #27559) [Link]

if you do not understand the security features of openbsd, you should not comment on them.

>> i've spent quite some years in computer security

yet you have never spent any time reviewing any of the information on openbsd in order to submit an informed post on the topic. i'm not saying that its featureset is not subject to criticism, but you aren't even doing that, you're just offering more hand-waving.

Performance v.s. Quality

Posted Nov 7, 2005 18:08 UTC (Mon) by PaXTeam (subscriber, #24616) [Link]

now where did you figure out that i "don't understand the security features of OpenBSD"? or that i "have never spent any time reviewing any of the information on OpenBSD"? did you read all that out of my post above? google up PaX one day then Mr. i-know-security-and-openbsd. sorry, i meant yahoo or whatever. once you do that and realize what it's been about for the past 5 years, you'll probably also understand why i can comment on their features (and have done so on numerous occasions, check bugtraq or undeadly.org).

also, you ditched my question so i'll ask it again: what is it exactly that allows a Linux or (say) NetBSD system to be compromised but doesn't allow it on OpenBSD? it's funny that you are asking me for submitting an informed post on OpenBSD security (not that this was the best forum for that, mind you), yet you fail do the same? you know what best describes you? your own words: your post is 100% content-free zealotry.

Performance v.s. Quality

Posted Nov 7, 2005 18:20 UTC (Mon) by b7j0c (subscriber, #27559) [Link]

>> what is it exactly that allows a Linux or (say) NetBSD system to be
>> compromised but doesn't allow it on OpenBSD?

openbsd is not designed to fix netbsd or linux security issues.

i never said netbsd or linux were insecure, you like every other poster here seems to only understand the pepsi-challenge mentality, that one wins and the other loses.

where did i say openbsd closes linux security holes?

where did i say linux has glaring security holes?

Performance v.s. Quality

Posted Nov 8, 2005 17:29 UTC (Tue) by PaXTeam (subscriber, #24616) [Link]

you said: "i would offer that openbsd is generally considered the most secure open source OS". if it's 'the most secure', then by definition the others are not the most secure, or in plain english, they're less secure. i asked what that 'less secure' is exactly and i have yet to receive an informed post from you. and in case you don't understand the word 'secure', i'll offer a simple definition: security is about information flow control, i.e., i expect you to give specific examples where OpenBSD provides information flow controls where the others don't (and 'provides' implies that such controls are bugfree, non-exploitable, else said controls are just an exercise in vain), and i also expect you to prove that all controls the 'less secure' systems have are also present in in OpenBSD (so that you can in the end prove that OpenBSD security is indeed a true superset of all the others). your turn sir ;-).

Performance v.s. Quality

Posted Nov 6, 2005 9:53 UTC (Sun) by rqosa (guest, #24136) [Link]

> these people use linux because of beowulf, not because it is "better" than freeBSD.

That statement makes no sense at all. From the Wikipedia article:

There is no particular piece of software that defines a cluster as a Beowulf. Commonly used parallel processing libraries include MPI (Message Passing Interface) and PVM (Parallel Virtual Machine). Both of these permit the programmer to divide a task among a group of networked computers, and recollect the results of processing.

It is entirely possible to have a "Beowulf" cluster running FreeBSD, or any other modern Unix, or even Windows.

> freeBSD is not a kernel, it is an operating system

That may have been true once, but now there are other distributions using the kernel of FreeBSD. This Wikipedia article lists the following: Freesbie, DesktopBSD, PC-BSD, PicoBSD, pfSense, m0n0wall, Gentoo/FreeBSD, and Debian GNU/kFreeBSD. Of course Linux has more distributions, but that's at least partly because it's more popular.

Performance v.s. Quality

Posted Nov 6, 2005 12:23 UTC (Sun) by samb (subscriber, #32949) [Link]

once again, better man pages, superior code delivery solution in /ports, and in my opinion, much better performance given heavy loads.

See, right now you are confirming what I wrote earlier about FreeBSD advocacy being about repeating memes.

The 'better performance under heavy load' is a claim, like that of 'better network performance' that goes back to the 90s. Again, it is something that was probably true at some time, but which probably hasn't been for many years.

Unless you have some benchmarks to show us? Considering that Linux is doing some very heavy lifting out there (kernel.org used to run with a load avg. of 200 back when the machine it was running on was a lot less beefy than today), finding itself in far more diverse environments than FreeBSD I doubt that you do.

Better man pages? Ok, that might be true for some of the bsd base tools vis-à-vis their GNU counterparts. I haven't compared. But if man pages is what you crave, you really should be using Debian, where it is a matter of policy to provide man pages for every package, even if they have to write it themselves. FreeBSD ports has no such policy.

Superior code delivery? I beg to differ. FreeBSD ports has nothing that matches the mandatory Debian Policy in this regard.

Man pages

Posted Nov 6, 2005 17:39 UTC (Sun) by Ross (subscriber, #4065) [Link]

I have to agree the three primary BSDs all have more complete and up to date man pages than Fedora or Debian.

However, the discussion was about the performance claims vs. Linux in the announcement, not about quality of documentation vs. Linux. :)

Performance v.s. Quality

Posted Nov 7, 2005 4:07 UTC (Mon) by b7j0c (subscriber, #27559) [Link]

>> Unless you have some benchmarks to show us? Considering that Linux is doing some very heavy lifting out there

my environment is the #1 most visited web site, which i think constitutes a useful case study.

Benchmarking

Posted Nov 7, 2005 4:15 UTC (Mon) by Ross (subscriber, #4065) [Link]

So have you done any benchmarking? Care to post the results?
Case studies are interesting too. Is there one for the site
in question? And, BTW, what Web site is it you are talking about?

Performance v.s. Quality

Posted Nov 7, 2005 4:19 UTC (Mon) by cventers (subscriber, #31465) [Link]

"#1 most visited web site" isn't much of an interesting metric without
context or convincing evidence that you have any connection to said
environment. You seem to be the "#1 most clueless poster" in this
discussion thread.

But I'll bite. Let's say you're right, and let's just say you're an
administrator for Yahoo, which may or may not be "#1 most visited
website" but is certainly high on the list. Consider that Yahoo has been
around since the days when your claims are true, and hence the staff are
BSD savvy. Why would they change to Linux unless BSD simply failed them?

And if we continue to assume that you do work at Yahoo, I'd like to point
out your competitor, Google, who uses Linux to power their supermassive
search network. Don't they constitute a useful case study too?

Performance v.s. Quality

Posted Nov 7, 2005 4:41 UTC (Mon) by b7j0c (subscriber, #27559) [Link]

>> You seem to be the "#1 most clueless poster" in this
>> discussion thread.

i'm not the one defaming people's character by name in a public forum.

>> Why would they change to Linux unless BSD simply failed them?

we also use redhat.

>> And if we continue to assume that you do work at Yahoo,
>> I'd like to point
>> out your competitor, Google, who uses Linux to power their supermassive
>> search network. Don't they constitute a useful case study too?

of course, linux is a powerful OS. i have not made any claims to the contrary.

once again, i use ubuntu linux. i like it. it is the first linux i have found that approaches the integrated system approach of freebsd.

A note to others

Posted Nov 7, 2005 19:26 UTC (Mon) by gravious (guest, #7662) [Link]

cventers, PaxTeam and others. I would not waste your breath replying to b7j0c. I have been bitten by this poster in the past. This perp is probably a random word generator with a bit of context thrown in. Browse any among
http://www.google.ie/search?hl=en&q=b7j0c+site%3Alwn....
to see what I mean. Honestly, save yer collective brain-cells for a worthier project.

I'll point out that b7j0c is a paying lwn.net subscriber

Posted Nov 8, 2005 0:17 UTC (Tue) by xoddam (subscriber, #2322) [Link]

So (s)he can't be all bad :-)

I'll point out that b7j0c is a paying lwn.net subscriber

Posted Aug 13, 2006 10:52 UTC (Sun) by gravious (guest, #7662) [Link]

Fair enough I suppose. I should fight the temptation to polarize everything. Easier said than done though.

Performance v.s. Quality

Posted Nov 6, 2005 21:06 UTC (Sun) by emkey (guest, #144) [Link]

>> haha are you serious? these installations are so tweaked they bear little resemblance to your Gentoo box.

You're showing a fair amount of ignorence here as another poster has already pointed out. In my experience (And yes, I do have have extensive experience with several Linux systems in the current top 500.) the only differences between the Linux stack on most of these systems and your typical desktop is the addition of drivers for high speed low latency interconnects such as those proveded by Quadrics, Myrinet and Infiniband vendors and the addition of MPI. These two items are intrinsic to HPC, not Linux. You'll find them on any vendors offering.

>> these people use linux because of beowulf, not because it is "better" than freeBSD.

Linux is used because it is widely supported, inexpensive and robust. How many high speed, low latency interconnects are supported in the various *BSD's?

As for Theo, I was on the netbsd list at that time. He was an abbrassive pain in the back side but he was also about the only guy getting anything done at the time so far as I could see. Personally I favored his side in that particular BSD splintering.

I recall David Miller, who did much of the porting of Linux to the Sparc architecture joining the NetBSD Sparc mailing list at one point and offering to help out. He was treated HORRIBLY. This and many other incidents soured me on the BSD camp forever.

I'd suggest to the BSD zeolots that they A, remove the chip from their shoulder and shutup, B, get coding and C, let nature take its course.

Performance v.s. Quality

Posted Nov 6, 2005 5:39 UTC (Sun) by tzafrir (subscriber, #11501) [Link]

b7j0c wrote:

> compare "make world" to updating facilities in any
> distro with perhaps maybe a nod to debian (which
> gets trashed for being too old-fashioned).

Practically any distro today has something similar to Debian's apt-get. At least for the functional equivalent of 'apt-get upgrade': a system-wide upgrade.

Also note that while FreeBSD may boast the good quality of the base system, most of the software is in the ports collection. Thus FreeBSD is as much a distribution as any Linux distro.

The repository size of the ports collection of FreeBSD is basically comparable to the ones of Debian and Gentoo: the major "community" distros. It is also interesting to see how Fedora's 'extras' have grown.

Performance v.s. Quality

Posted Nov 7, 2005 4:47 UTC (Mon) by b7j0c (subscriber, #27559) [Link]

>> Practically any distro today has something similar to Debian's apt-get.

"make world" is not the equivalent of apt/dpkg

it is a full operating system rebuild tool

in this sense the closest analogy is a kernel recompile mixed with apt-get dist-upgrade (but without having to change /etc/apt.sources etc)

Performance v.s. Quality

Posted Nov 7, 2005 14:54 UTC (Mon) by Los__D (subscriber, #15263) [Link]

You want that for a Linux?

Here's a script for you for Gentoo's portage, an improved (*10) ports:

#!/bin/sh
emerge -uDv system && emerge -uDv world && genkernel --menuconfig

There's your 'make world' all in one command...

Very important to write ONE command, I can see that, don't want too much flexibility or anything!

Performance v.s. Quality

Posted Nov 7, 2005 14:56 UTC (Mon) by Los__D (subscriber, #15263) [Link]

Remember to throw in an 'emerge sync' first if you don't want choice of upgrading the tree as well...

Performance v.s. Quality

Posted Nov 7, 2005 17:04 UTC (Mon) by b7j0c (subscriber, #27559) [Link]

yes i understand that other systems have these features. freeBSD has had them for nearly a decade, that is the difference. freeBSD had these features before Gentoo existed.

Gentoo is an excellent linux distribution and offers its users many of the quality configiruation utilities that freeBSD users have enjoyed for a very long time.

Performance v.s. Quality

Posted Nov 7, 2005 17:59 UTC (Mon) by dlang (✭ supporter ✭, #313) [Link]

but nobody is disagreeing that *BSD was better then Linux a decade ago. everyone but the most fanatic linux folks (who also have to be ignorant of history), many people would probably say that *BSD was better then Linux 5 years ago (but this would be a smaller number), the question at hand is how do they compare today.

the rate of change of Linux is such that (assuming it's improving things) Linux will be better then *BSD at some point, the only real question is if that point has been reached yet.

Performance v.s. Quality

Posted Nov 10, 2005 10:18 UTC (Thu) by ekj (subscriber, #1524) [Link]

Okay. So let's compare.

Mandriva: "urpmi.update -a" to update the lists of available packages, thereafter "urpmi --auto-select" to select and install upgraded packages. Dependencies are handles automatically (like for all of these systems really)

Fedora: "yum update"

Debian: "apt-get update && apt-get upgrade"

I don't see the point. All modern Freenixen has package-management and dependency-resolution, arguably the "make world" concept from BSD is measurably inferior in a lot of ways. How do you do any of the following with the "make world" approach:

Efficiently upgrade a machine with no installed development-environment.
Upgrade a machine with very little CPU power without waiting a month.
Answer the question: "/foo/bar is a strange file, which package put that there?"
Answer the question: "If I wanted to uninstall python, what else would need to go?"
Answer the question: "What is the total installed size of OO ?"

All of which are trivial in any of my mentioned package-management systems, and as far as I know distinctly nontrivial with a "make world" based system of clever makefiles.

Performance v.s. Quality

Posted Nov 15, 2005 12:21 UTC (Tue) by hppnq (subscriber, #14462) [Link]

Not to mention that you probably don't want your package management to be horribly broken by the fact that you've lost contact with your NTP server.

== Cathedral v.s. Bazar

Posted Nov 10, 2005 21:21 UTC (Thu) by ernest (subscriber, #2355) [Link]

Which is probably the only reason why GNU/Linux got so big.

Ernest.

Performance v.s. Quality

Posted Nov 5, 2005 22:28 UTC (Sat) by samb (subscriber, #32949) [Link]

FreeBSD produces a high quality UNIX operating system by following these principles.

I find it odd that anyone can have such an idealized view of FreeBSD development considering what happened with 5.x. Delayed by several years, and when it was finally released it was still not ready for prime time wrt what had been its main targets, SMP and multithreading, issues that are now being adressed in 6.x.

Things looked very different back when CURRENT was 4.x. Linux took some large leaps forward while FreeBSD struggled with 5.x.

They have had some tough times

Posted Nov 5, 2005 22:39 UTC (Sat) by b7j0c (subscriber, #27559) [Link]

The 5 series was unfortunate, like the 3 series.

They had some central figures move on.

They had to wrestle with distractions like DragonFly.

They had to do some major code changes.

Many of us simply kept going with 4.x (a fantastic release) while these issues were worked out.

Some of the better kernel people I know are saying 6.x is going to be a solid release that bolsters BSD's good reputation.

FreeBSD Project Launches FreeBSD 6.0

Posted Nov 6, 2005 4:49 UTC (Sun) by nurhussein (guest, #16226) [Link]

Gentlemen, please.

The claim in the press release was this:

Recent performance benchmarks show that FreeBSD 6.0 outperforms Linux in raw data throughput.

I've heard many similar claims of superior performance throughout my interactions with FreeBSD afficionados, but for the purpose of this discussion let us just look at the above claim. I'm curious as to where these benchmarks are, how they were conducted and where they were published.

If other operating systems are to gain anything from the FreeBSD experience, we need to know and why FreeBSD is (reportedly) getting better performance over, say, Linux. But without adequate information, we can't do that, now can we.

random-io benchmarks: FreeBSD Project Launches FreeBSD 6.0

Posted Nov 17, 2005 3:51 UTC (Thu) by NuShrike (guest, #33902) [Link]

My recent benchmarks at work shows that FreeBSD 6.0-stable seems slower than FreeBSD 5.4-stable on the same hardware for random-io.

I used rawio from Ports modified to allow file-reads. It is a custom kernel build from the same conf file.

Compaq DL380 G2 with 2 Pentium III 1.4GHz 512K L2, 2GB ram, 2 36GB raid-1s (OS /, data /usr2)
/dev/da1s1d on /usr2 (ufs, local, noatime, soft-updates)

I run each test 3 times, and take the third result
flushram allocates 1024*1024*1850 of ram before it quits

FreeBSD 5.4-stable
GCC 3.4.2 (gcc -O2)

./FBSD.flushram ; ./FBSD.rawio -v 2 -p 1 -F -f -c 4096 /usr2/blah/blah.dat
Test ID Time KB/sec /sec %User %Sys %Total Reads Writes
RR blah.dat.2048 54.035362 1241.9 303 0.0 1.9 1.9 16384 0
SR blah.dat.2048 0.196668 341229.2 83308 0.0 100.0 100.0 16384 0

FreeBSD 6.0-stable
GCC 3.4.4 (gcc -O2)

./6FBSD.flushram ; ./6FBSD.rawio -v 2 -p 1 -F -f -c 4096 /usr2/blah/blah.dat
Test ID Time KB/sec /sec %User %Sys %Total Reads Writes
RR blah.dat.2048 57.725636 1162.5 284 0.1 1.9 2.0 16384 0
SR blah.dat.2048 0.198070 338813.9 82718 7.4 92.6 100.1 16384 0

FreeBSD Project Launches FreeBSD 6.0

Posted Nov 6, 2005 7:40 UTC (Sun) by error27 (subscriber, #8346) [Link]

Yeah. Maybe there is a benchmark somewhere where FreeBSD kicks Linux's butt. It would be interesting to see what that benchmark was. Generally through put is a hardware/driver issue not a block layer or network stack issue these days.

Benchmarks are just a starting point. For example, if you say that application start up times in Linux are bad then people can verify it and look at what's causing the problem.

This claim is meaningless so that probably means the person doesn't know what he's talking about.