Gentoo Forums :: View topic - Roy is gone... Should I Leave Gentoo

Roy is gone... Should I Leave Gentoo

View unanswered posts
View posts from last 24 hours

Goto page Previous 1, 2, 3, 4, 5

Gentoo Forums Forum Index

Gentoo Chat

View previous topic :: View next topic

What OS Should I Use

Stick With Gentoo: Things will straighten out

73%

[ 126 ]

FreeBSD

[ 12 ]

Fedora

[ 2 ]

Debian

[ 1 ]

Ubuntu

[ 9 ]

Sabayon

[ 3 ]

Stick With Gentoo: Switch to Paludis and other 3rd party overlays

[ 14 ]

Another Distro

[ 4 ]

Total Votes : 171

Author

Message

vermaden
Apprentice

Joined: 27 Jul 2005
Posts: 168
Location: pl_PL.lodz

Posted: Wed Nov 28, 2007 7:56 am Post subject:

Simba7 wrote:

How long have you been using FreeBSD, vermaden?

From 5.4 in middle 2005, before that I used Gentoo for about a year and Slackware for about a year maybe, something like that.
_________________
"If 386BSD had been available when I started on Linux, Linux would probably never had happened." Linus Torvalds
vermaden's: links spreadbsd

Matteo Azzali
Retired Dev

Joined: 23 Sep 2004
Posts: 1133

Posted: Wed Nov 28, 2007 3:33 pm Post subject:

vermaden wrote:

Matteo Azzali wrote:

Also, saying that it's time to move to BSD because of 10 slides in pdf
babbling about a rewrited system.... :roll:

(makes me think about HURD).
Maybe in 2-3 years when that rewrited kernel will start to be stable
the thing could be reconsidered, not actually, and surely not for users.

This statement shows how small your knowledge about FreeBSD is, and that you never had ANY experiences with this UNIX.

Rewrited system? HURD? Please, did we read the same PDF file?

FreeBSD 7 has been STABLE for months, providing performance, scalability and reliability that Linux can currently only dream of.

Surely not for users? Dont tell me that you recommend Gentoo for users, dont you?

Have you forgotten you roots soldier? All Gentoo concepts and design are based on FreeBSD's design, Portage = slower Ports, init system = FreeBSD's RC (imported from NetBSD), USE flags = WITH_*/WITHOUT_* ...

You generally do not understand that bashing FreeBSD you are curently bashing Gentoo.

If FreeBSD would not exist then Gentoo would never happen.

Nobody forces you to use FreeBSD, but stop spreading FUD if you dont know shit.

1) Those slides about a complete kernel rewrite I was talking about are 20/oct/2007 ones and are PREVIEW,
(read under the title...) and there are zero benchmarks on the net about the claims there.
And still, in OSS even marking stable (thing that FreeBSD folks haven't done!!!) the project wouldn't mean
it's really stable (you can search the kde 4 discussions and delayed release plan if you don't understand).
Also note that the release process of FreeBSD7 started in june 2007 (less than six months ago)

2) I'm not bashing anything, someone showed some slides about a rewritten FreeBSD as the Holy Graal.
I just don't believe in holy graals, and in those marked stable too fast I believe even less, call me stupid!!!
Give proofs, not words, point us a link with a trustable source making benchmarks and comparisons,
otherwise you're just doing propaganda in the gentoo forums (what if I come in FreeBSD forums
saying that gentoo is more modular, has a better package managment and not only portage tree is bigger
than any distro official repos, but with ovelays gentoo has more than twice the packages than FreeBSD??)....

Beside that, I admit that I don't use FreeBSD and last time I used it there wasn't even an X environment,just pine
and not much more,so you can even tell that FreeBSD creates nice gals for it's owners pleasure and I can't debunk
(but I would still think that likely those girls are hallucinations , for Occam's razor)....
_________________
Every day a new distro comes to birth. Every day a distro "eats" another.
If you're born distro, no matter what, start to run.
---- http://www.linuxprinting.org/ ---- http://tuxmobil.org/

Matteo Azzali
Retired Dev

Joined: 23 Sep 2004
Posts: 1133

Posted: Wed Nov 28, 2007 4:04 pm Post subject:

Oh, just because I didn't mentioned that before, I heard complains about overlays and things staying out of the tree. Please show me which distro (linux or bsd) has more than 120'000 packages in the official repositories (as portage tree is the official repo, while overlays= unhofficial repos). _________________ Every day a new distro comes to birth. Every day a distro "eats" another. If you're born distro, no matter what, start to run. ---- http://www.linuxprinting.org/ ---- http://tuxmobil.org/

Simba7
l33t
l33t

Joined: 22 Jan 2007
Posts: 706
Location: Billings, MT, USA

Posted: Wed Nov 28, 2007 4:52 pm Post subject:

vermaden wrote:

Simba7 wrote:

How long have you been using FreeBSD, vermaden?

From 5.4 in middle 2005, before that I used Gentoo for about a year and Slackware for about a year maybe, something like that.

Man, I've been using FreeBSD since 3.0 back in 1998. Gentoo more recently in 2005. I like 'em both, but I'm starting to gain more interest in Gentoo due to support on some of my hardware. It's nice to see something that works natively (no NDIS wrapper) instead of being told "Use a different card".

What bothers me with *BSD and Linux is that some users have a huge ego trip on which one is better without trying the other. This is just stupid. How do they know if it's solid or not if they don't try it? This is another thing that turned me away from Linux.

Gentoo isn't your typical Linux, though. Probably why I enjoy using it.

vermaden
Apprentice

Joined: 27 Jul 2005
Posts: 168
Location: pl_PL.lodz

Posted: Wed Nov 28, 2007 5:41 pm Post subject:

Matteo Azzali wrote:

1) Those slides about a complete kernel rewrite I was talking about are 20/oct/2007 ones and are PREVIEW, (read under the title...)

Complete kernel rewrite? Are you sure what are you talking about? There is page about new architecture ARM and only there I saw a rewrite keyword, and not the kernel, exacly this is the quote:
"support for Avila Gateworks Xscale boards was added, including a rewrite of the Intel code"

FreeBSD 7 was focused about scalability and SMP performance, with SMPng project [ http://freebsd.org/smp ], they rewrited only FOUR subsystems, and most in 2000.

Code:

YEAR what(arch)
2000 Rewrite the low level interrupt code (i386 UP)
2000 Rewrite the low level interrupt code (i386 SMP)
2000 Rewrite the low level interrupt code (alpha)
2002 Rewrite kernel memory allocator to be a slab allocator that uses per-cpu caches
2002 Rewrite kernel memory allocator so that Giant is not required for malloc() or free()

Maybe you got some other document where it is cleanly said that they rewrited the whole kernel thus year? FreeBSD developers would be very surprised about that also.

Matteo Azzali wrote:

and there are zero benchmarks on the net about the claims there.

On which Internet you were searching? Propably not the one that I use:
http://people.freebsd.org/~kris/scaling/4cpu-mysql.png
http://people.freebsd.org/~kris/scaling/4cpu-pgsql.png
http://people.freebsd.org/~kris/scaling/linux-mysql.png
http://people.freebsd.org/~kris/scaling/linux-pgsql.png
http://people.freebsd.org/~kris/scaling/mysql-freebsd.png
http://people.freebsd.org/~kris/scaling/os-mysql.png
https://manuelmartini.it/bench/mysql/
http://blog.insidesystems.net/articles/2007/04/11/postgresql-scaling-on-6-2-and-7-0
http://www.netbsd.org/~ad/sysbench/sysbench-4bsd.png
http://www.netbsd.org/~ad/sysbench/sysbench-pcpu.png
http://people.freebsd.org/~jeff/mysqlwrite.png
http://www.netbsd.org/~ad/sysbench2/4cpu.png
http://jeffr-tech.livejournal.com/5705.html
... should I for google more?

Matteo Azzali wrote:

2) I'm not bashing anything, someone showed some slides about a rewritten FreeBSD as the Holy Graal.

post links here.

Matteo Azzali wrote:

but with ovelays gentoo has more than twice the packages than FreeBSD??)....

FreeBSD has curently about 17000 ports aviaalble, debian about 22000, (but they like to split packages into smaller pieces, headers, modules, common, configs, etc), but personally I have never had instaled myself more then 700-800 ports, while about 300 goes for modular xorg ...

Matteo Azzali wrote:

Beside that, I admit that I don't use FreeBSD and last time I used it

Maybe you should?

Matteo Azzali wrote:

Please show me which distro (linux or bsd) has more than 120'000 packages in the official repositories
(as portage tree is the official repo, while overlays= unhofficial repos).

Does Fluxbox 1.0 + Fluxbox SVN + old Fluxbox 1.1.4 + Fluxbox 9.15 + Fluxbox 1.0RC2 counts as FIVE in all these trees? If yes ,then we can have many many trees and this will make total number of ports = trees * 17000, why not, we can do that.

Simba7 wrote:

Man, I've been using FreeBSD since 3.0 back in 1998. Gentoo more recently in 2005.

Nice.

Simba7 wrote:

What bothers me with *BSD and Linux is that some users have a huge ego trip on which one is better without trying the other. This is just stupid. How do they know if it's solid or not if they don't try it? This is another thing that turned me away from Linux.

This is sad and jeluous, but its unformatunely true also.

Simba7 wrote:

Gentoo isn't your typical Linux, though. Probably why I enjoy using it.

That is biggest advantage of it I think, fresh/other/bsd way of doing things, like Arch Linux for example, also nice attitude.
_________________
"If 386BSD had been available when I started on Linux, Linux would probably never had happened." Linus Torvalds
vermaden's: links spreadbsd

Matteo Azzali
Retired Dev

Joined: 23 Sep 2004
Posts: 1133

Posted: Wed Nov 28, 2007 7:36 pm Post subject:

@vermaden: Ok I was misguided by pdf page 4 :

kris in the pdf wrote:

Goal: Re-design the FreeBSD kernel as a multi-threaded system,
for “next generation” SMP support

And the only improvements are, in reality , 4 modules (subsystems) modified in the year 2000. This is my error but it stops here.

vermaden wrote:

.....

Just compare the first graph of the pdf you posted : http://people.freebsd.org/~kris/scaling/7.0 Preview.pdf
(page 11 or page 15) with the same link you posted above :http://www.netbsd.org/~ad/sysbench2/4cpu.png
(ah, with reliable source I was referring to someone else, not FreeBSD and netBSD tests,
and less than ever kris tests as kris is the same author of that pdf, however I'll just use those....)

And... TA DA! The pdf shows performances more than doubled respect FreeBSD 6.0 (both in Mysql and postgreSQL),
While the same link you provided (from netBSD, still not slashdot but at least is another source) shows
about 20% (notice the vertical axis is starting from 100 and not from zero) more performance than plain linux and
about 5% less performances than linux + tcmalloc (which is not really optional for who uses SQL databases intensively,
on linux, as these machines do usually SQL only* them are equipped with tcmalloc,hoard or nedmalloc).
For the chronicles, the 2 netbsd links (one and two) you posted show netBSD faster than FreeBSD 7.....
while the PDF you linked showed completely different results in the same tests in pages 17/18.
(that's why I ask for external sources.... :roll:

)
Have you read the links you posted???!?!?
Them aren't the proof that the pdf we're talking about is thrustworthy, all the opposite.

So reality is that FreeBSD 6.0 was really bad, while 7.0 is now on pair with netBSD and the likes????!?

And where are all those improvements? How can the pdf author say "linux 2.6.22 is still 15% slower than FreeBSD 7" (without pointing
out that it's just in sql, just with 8+ threads and just if the linux admin forgot to use tcmalloc/hoard/nedmalloc....) ???? It's at page 19.

vermaden wrote:

Maybe you should?

I don't use sql intensively, and if I would I could still try to get the best performances with linux + tcmalloc
(I'm on a 32-bit compiled system and can't use "-fomit-frame-pointer" anyway cause I need decent backtraces).
But thank you.

*As general purpose, glib/gcc standard alloc is reported to be better or on pair,
and nothing is said about FreeBSD 7 (alloc or whole system) in general purpose
(both the pdf and the tests you linked talk just about sql and nothing more, so there's no way to know
if it's an improvement just in *SQL like tcmalloc/hoard/nedmalloc).
_________________
Every day a new distro comes to birth. Every day a distro "eats" another.
If you're born distro, no matter what, start to run.
---- http://www.linuxprinting.org/ ---- http://tuxmobil.org/

vermaden
Apprentice

Joined: 27 Jul 2005
Posts: 168
Location: pl_PL.lodz

Posted: Wed Nov 28, 2007 11:21 pm Post subject:

Matteo Azzali wrote:

So reality is that FreeBSD 6.0 was really bad, while 7.0 is now on pair with netBSD and the likes????!?

These are not results from the same time, for example NetBSD team wanted to compare their results with SMP rewrite, and they used FreeBSD with ALL debuging enagled which lead to 400% performance decrease in some areas, also after they compared results for 8 cores, they realized that FreeBSD is almost two times faster then NetBSD, and that they still have a lot of work.

You can wait for FreeBSD 7.0-RELEASE and benchmarks made then.

TCMalloc done is a great inproovement for Linux but it also have disadvantages, from TCMalloc site:
TCMalloc may be somewhat more memory hungry than other mallocs, though it tends not to have the huge blowups that can happen with other mallocs. In particular, at startup TCMalloc allocates approximately 6 MB of memory. It would be easy to roll a specialized version that trades a little bit of speed for more space efficiency.

TCMalloc currently does not return any memory to the system.

Don't try to load TCMalloc into a running binary (e.g., using JNI in Java programs). The binary will have allocated some objects using the system malloc, and may try to pass them to TCMalloc for deallocation. TCMalloc will not be able to handle such objects.

So its also not ready for production use, but after some time it should be.

Matteo Azzali wrote:

MySQL/PostresSQL benchmarks are not the only place where FreeBSD or Linux are great performers, but its easy to use SQL benchmarks to meassure overall system performance since it is very demanding on a lot of subsystems, and is good helper to find hot places in code.

Matteo Azzali wrote:

and nothing is said about FreeBSD 7 (alloc or whole system) in general purpose

FreeBSD 7 uses jemalloc: http://people.freebsd.org/~jasone/jemalloc/bsdcan2006/jemalloc.pdf

Also you can contact Kris or Jeff and tell them which options/malloc tunables they should set/use to get better performance, they asked for this many times if Linux users posted that they used non tuned Linux.
_________________
"If 386BSD had been available when I started on Linux, Linux would probably never had happened." Linus Torvalds
vermaden's: links spreadbsd

Matteo Azzali
Retired Dev

Joined: 23 Sep 2004
Posts: 1133

Posted: Thu Nov 29, 2007 11:16 am Post subject:

vermaden wrote:

You can wait for FreeBSD 7.0-RELEASE and benchmarks made then.

That's what I say from my first post, too early to say what's better than what, too early to
blindly believe in a pdf + some benchmarks done by a much involved part
(Kris, again.....).
More, actually we pointed out that the improvements are for 4+ cpus systems
(8 threads+) and in SQL only.

Quote:

....
TCMalloc done is a great inproovement for Linux but it also have disadvantages, from TCMalloc site:
TCMalloc may be somewhat more memory hungry than other mallocs, though it tends not to have the huge blowups that can happen with other mallocs. In particular, at startup TCMalloc allocates approximately 6 MB of memory. It would be easy to roll a specialized version that trades a little bit of speed for more space efficiency.

TCMalloc currently does not return any memory to the system.

Don't try to load TCMalloc into a running binary (e.g., using JNI in Java programs). The binary will have allocated some objects using the system malloc, and may try to pass them to TCMalloc for deallocation. TCMalloc will not be able to handle such objects.

So its also not ready for production use, but after some time it should be.

Matteo Azzali wrote:

and nothing is said about FreeBSD 7 (alloc or whole system) in general purpose

Again, TCmalloc and the likes are used only when you badly need SQL performances, I stated clearly that for general
purpose it's not the best, but if you're running an SQL server:

a) you can avoid to start an application with malloc and then pass it to TCmalloc, you just start it directly with TCmalloc.
(let me say that's quite a stupid case. You start SQL with TCmalloc and everything else with malloc. period.
This doesn't means is not ready for production, means people must turn on their brain using it!!!!!!)
Also please check that there are incompatibilities even between jemalloc and phkmalloc, it's nothing so strange
that TCmalloc can't deallocate what malloc allocated.

b) 6 Mb of memory more "once, at startup" aren't much of an issue today (in y 2k7 we have 512-1Gig ram and some Ghz,
even if people still do benchmarks with pentium III that's from the past millennium , I swear why :roll:

)

About Jemalloc, you're again blindly following the tests from the BSD devs. And my question is still the same:
external source "super partes"? Again,even in those conditions, in the same Jeff tests dlmalloc is faster than Jemalloc
in most non-SQL tests (page 8 and 10), (and there's no test against glibc malloc in general purpose, at all....).

Quote:

????False. SQL is a specific task and doesn't measures overall system performance at all. As you can see in many
benchmarks on the net, there are allocator badly optimized for SQL that badly fail in all or most other fields.
(take hoard/nedmalloc/malloc for example).
MySQL is choosed because is one of the most power hungry (and adopted) application in the server field. And it has
not very much to do with desktop systems (even when you use amarok with mySQL, you'll never reach sizes for which
you can see the difference even if you're the amazon shop....)
_________________
Every day a new distro comes to birth. Every day a distro "eats" another.
If you're born distro, no matter what, start to run.
---- http://www.linuxprinting.org/ ---- http://tuxmobil.org/

Last edited by Matteo Azzali on Thu Nov 29, 2007 11:52 am; edited 1 time in total

UberLord
Retired Dev

Joined: 18 Sep 2003
Posts: 6835
Location: Blighty

Posted: Thu Nov 29, 2007 11:46 am Post subject:

Matteo Azzali wrote:

b) 6 Mb of memory more "once, at startup" aren't much of an issue today (in y 2k7 we have 512-1Gig ram and some Ghz,
even if people still do benchmarks with pentium III that's from the past millennium , I swear why :roll:

)

Embedded devices is where OSS is winning a lot right now.
Memory and processor speed are at a premium there, so yes that 6mb is very important.
_________________
Use dhcpcd for all your automated network configuration needs
Use dhcpcd-ui (GTK+/Qt) as your System Tray Network tool

Matteo Azzali
Retired Dev

Joined: 23 Sep 2004
Posts: 1133

Posted: Thu Nov 29, 2007 11:52 am Post subject:

UberLord wrote:

Matteo Azzali wrote:

)

Embedded devices is where OSS is winning a lot right now.
Memory and processor speed are at a premium there, so yes that 6mb is very important.

Yes, but you use embedded devices to do SQL intensive tasks ????!???
ARM is approaching the 700Mhz speed, but it's not what you'll find in custom devices which are
usually clocked at 400Mhz or less.
Even worse, SQL is a memory hungry task, requiring lots of Mb, and as you said that's no good
in embedded (ant likely that's why there are not much benchmarks of SQL on ARM)
Again, TCmalloc is needed only for SQL servers
(in everything else, malloc is more than enough and likely faster than jemalloc)
_________________
Every day a new distro comes to birth. Every day a distro "eats" another.
If you're born distro, no matter what, start to run.
---- http://www.linuxprinting.org/ ---- http://tuxmobil.org/

UberLord
Retired Dev

Joined: 18 Sep 2003
Posts: 6835
Location: Blighty

Posted: Thu Nov 29, 2007 12:12 pm Post subject:

No, but your point by itself made no mention of SQL or other such stuff - you just discarded low memory and slow systems as stuff from the past millenium and thus should not be used in benchmarks. I disagree with that as a lot of todays software is too bloated and slow instead of becoming more efficient and leaner. I also think that you only get good benchmark analysis on machines old and new, slow and fast so you get a good feel of how a program really runs in a mixture of environments. Sure, app A maybe faster than app B if the machine has 16gig of memory, but if app B is faster when said machine has 1gig of memory then I know which one I'd be using. _________________ Use dhcpcd for all your automated network configuration needs Use dhcpcd-ui (GTK+/Qt) as your System Tray Network tool

Matteo Azzali
Retired Dev

Joined: 23 Sep 2004
Posts: 1133

Posted: Thu Nov 29, 2007 12:20 pm Post subject:

UberLord wrote:

....
I also think that you only get good benchmark analysis on machines old and new, slow and fast so you get a good feel of how a program really runs in a mixture of environments.
.....

It's not just about faster, is about comparisons. Different approaches/algorithms have different results based
on technologies improvements made by cpus.
i.e.: really different cache size may completely change the figures in a lot of tasks, based on the
algorithm, and pentium III 700Mhz had 256 kb of L2 cache and 32 Kb of L1, while Core 2 Duo for example
have 64 Kb of L1 and more than 1 Mb (usually 1-2 Mb, so 4-8 times pentium III) L2 per core!!!!!
And this cache growth has been proved being a big performance increase with a lot of algorithms (but not all).
The same applies for memory speed, the differences between access times DO change the performances
based on which algorithm you're using.

That's why I think that algorithms/software comparisons benchmarks should be done on system at least near* to those
on which you'll run the algorithm, or them are worth nothing.

*You may consider a core duo , which is not really anymore on the market, as "near" to a core 2 duo, and
expect that the results are at least similar. Pentium D, Pentium IV and pentium III are instead quite different,
and their multi-cpu configuration has really different timings from multi core processors, being forced to
use an external bus. More than this, if you get Pentium IV and Pentium D, you have hyperthreading which
has almost unpredictable performances effect depending on the optimization of your code for such technology.
And in january we'll have Penrin and Phantom, which will likely change the figures another time.....

If you're still skeptic, get a benchmark comparison of Q6600 in the two flavours: old stepping and G0,
there are performance increases bigger in some tests and lower or null in some others. (from 0 to 5%
depending on the software). And all that just changing the stepping with the same process ,the same
technologies, cache size, memory and cpu speed!!!!!
_________________
Every day a new distro comes to birth. Every day a distro "eats" another.
If you're born distro, no matter what, start to run.
---- http://www.linuxprinting.org/ ---- http://tuxmobil.org/

Display posts from previous:

	Gentoo Forums Forum Index Gentoo Chat	All times are GMT Goto page Previous 1, 2, 3, 4, 5
Page 5 of 5

Jump to:

You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum

Copyright 2001-2024 Gentoo Foundation, Inc. Designed by Kyle Manna © 2003; Style derived from original subSilver theme. | Hosting by Gossamer Threads Inc. © | Powered by phpBB 2.0.23-gentoo-p11 © 2001, 2002 phpBB Group
Privacy Policy