Re: [opensuse] SUPER, SUSE Performance Enhanced Release.
3.) Personally I want to stay as close as possible to the openSUSE code, since this means that each openSUSE release gets followed on
day by our "addon" packages. ... In short I do not want to create a new distro out of this or a fork
That's a most commendable attitude. As I said above, this way I get
Hello there I send my reply to the list, since I might not have been clear enough about one particular issue. the the
best of both worlds!
Those packages should and could be installed on top of an existing openSUSE install
I strongly support this idea, since this will enable ordinary users
Correct, Compatibility is certainly what I am personally after. What some of us want to do additionally is really up to the project, but personally I want things to just be usable by any standard openSUSE user by pointing the system out to a YOU/Yum or apt repository, containing a set of packages that will replace your current ones with optimised ones. Same code, just tweaked to be faster. What the project will come up with is really not just up to me, but I certainly do have my views on this matter, which are inline with yours. like
me to install the enhancement too.
we could also create a 1 CD base desktop distro a user can download and install quickly and then install additional packages onto with Yast or Kynaptic/Synaptic against the gwdg.de repository ,
This would mean difficulty for me, since I do not have access to unlimited bandwidth, I can't download everything myself. Right now I
have about 4 gig of SuSE 9.3 installed on my system, and all but 200 KB of it is from the CDs. So I really prefer the idea with the packages on top of a clean SUSE 10 install.
really like. We could also create a mini install containing <150
packages on top of which a geek can install the rest. A real minimal install, which I personally also like.
Ah, the keyword here is "geek". We are not geeks, but still would
I probably did not explain it well enough, but I was actually talking about at least 2 iso's. The #1 iso is already a full desktop. You get a full blown base desktop (KDE) and then just add things you want additionally. The mini-install is just for me and my requirements to have a minimal mini server install. Something that could be further optimised to create one of those USB stick distro's ;). If anyone wants to add a 3rd iso (Gnome Desktop), then anyone can do that. Thanks to the awesome work the guys and girls in Nuernberg and Prag do and did, this is all very easy with SUSE. base like
to take the advantage of your tweaking. So such a mini install would not be for us.
I think the iso #1 is for the home user or anyone really. 1 CD comes with all you need to get a desktop going and installs without asking too many questions (using autoyast.xml). The mini geek CD is just for me and Mike and others who like the rough and small since we both like to have a very small Linux running on some of our systems.
5.) Personally, we should have scripts at the end of this which automatically create our custom version out of each SUSE release. That we we can spend time fine tuning instead of constant hacking and packaging.
I hope that this custom version will be the additional packages like you said above, and not a separate release. I mean, not even like NLD is
different from SUSE.
Yeah it will be IMHO like an enhancement pack to your normal distro installable by anyone via YOU/Yum/Apt. Once we have scripts ready we could potentially run the same scripts against SLES/NLD and create an optimised version of those. Well so the theory. At least I can get a few of my colleagues in the office happy.
In short, my mail is to request you to not forget the home user who is *not* a geek, and allow him to run SUSE on steroids anyway!
Always has been my intention, and I probably did not explain it well enough, thanks for asking and allowing me to explain this more. The geek comment was referring to iso #2, the mini install. Andreas
Andreas Girardet wrote:
I probably did not explain it well enough, but I was actually talking about at least 2 iso's. The #1 iso is already a full desktop. You get a full blown base desktop (KDE) and then just add things you want additionally. The mini-install is just for me and my requirements to have a minimal mini server install.
Ah I see. There are many aspects to your proposal: 1. create optimized parts of the OS that we (ordinary non-geek users) can download as "updates" - this is for people like me who want a full fledged SUSE 10 install, but running at top performance. 2. create a single CD install version of SUSE 10 with optimization and common applications (on KDE, yay, my fav desktop!) - this is for giving people a "taste" of SUSE (correct me if I'm wrong), but SUSE running hyper (or maybe hi-per = high performance) 3. create an optimized mini-install that is enough for servers etc - this is for people like you who are geeks who need a basic OS only on specialized machines like file/web/mail servers etc. But I have a basic question: If you are able to do these optimizations, then why not integrate them into the main software itself? Why make it a separate downloadable package? Let SUSE *by default* run at high performance. What's the downside to that? BTW regarding creating multiple "renditions" of the same SUSE packages, I think you could have a glance at the "Should we use jigdo for some kind of downloads?" thread. -- Shriramana Sharma http://samvit.org
Shriramana Sharma wrote:
If you are able to do these optimizations, then why not integrate them into the main software itself? Why make it a separate downloadable package? Let SUSE *by default* run at high performance. What's the downside to that?
Thats the whole point. Providing proof of concepts and doing the leg work for SUSE/Novell. I can bounce up and down and off the walls about some wiz-bang foo application, but it takes a great deal of time to a) package b) build c) test d) test e) test some more. The idea is we build and package the wiz-bang application and or patch sets, prove stability and that it does infact provide performace enhancements. Thus makingthe path of acceptence into the SUSE/Novell develoment tree an easier one. Mike
On 8/15/05, Michael Honeyfield
Thats the whole point. Providing proof of concepts and doing the leg work for SUSE/Novell.
I can bounce up and down and off the walls about some wiz-bang foo application, but it takes a great deal of time to a) package b) build c) test d) test e) test some more.
The idea is we build and package the wiz-bang application and or patch sets, prove stability and that it does infact provide performace enhancements.
Thus makingthe path of acceptence into the SUSE/Novell develoment tree an easier one.
Mike
I really think this is a great idea. One of the things that I do not like about SUSE is the speed. FC is faster, but not as well rounded. I would love to help with getting SUSE up to speed. -- Ken Swain ken@kenswain.com http://kenswain.com "Spoof them all, let god sort them out."
Who's organizing the Super SuSE effort?
--- Ken Swain
Thats the whole point. Providing proof of concepts and doing the leg work for SUSE/Novell.
I can bounce up and down and off the walls about some wiz-bang foo application, but it takes a great deal of time to a) package b) build c) test d) test e) test some more.
The idea is we build and package the wiz-bang application and or patch sets, prove stability and that it does infact
On 8/15/05, Michael Honeyfield
wrote: provide performace enhancements.
Thus makingthe path of acceptence into the SUSE/Novell develoment tree an easier one.
Mike
I really think this is a great idea. One of the things that I do not like about SUSE is the speed. FC is faster, but not as well rounded. I would love to help with getting SUSE up to speed.
-- Ken Swain ken@kenswain.com http://kenswain.com "Spoof them all, let god sort them out."
---------------------------------------------------------------------
To unsubscribe, e-mail: opensuse-unsubscribe@opensuse.org For additional commands, e-mail: opensuse-help@opensuse.org
__________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com
Thanks Mike. Have we broken down whats needed to get
the project rolling? I think we need to take
inventory of people 'in' on the project and related
skills.
--- Michael Honeyfield
Carlos wrote:
Who's organizing the Super SuSE effort?
--- Ken Swain
wrote: If you check through this thread, Andreas of Novell NZ is kick starting it.
Mike
---------------------------------------------------------------------
To unsubscribe, e-mail: opensuse-unsubscribe@opensuse.org For additional commands, e-mail: opensuse-help@opensuse.org
____________________________________________________ Start your day with Yahoo! - make it your home page http://www.yahoo.com/r/hs
No, it was only announced yesterday. Email me (or Andreas) off list and I will add your name if you are interested. Cheers Mike Carlos wrote:
Thanks Mike. Have we broken down whats needed to get the project rolling? I think we need to take inventory of people 'in' on the project and related skills.
--- Michael Honeyfield
wrote: Carlos wrote:
Who's organizing the Super SuSE effort?
--- Ken Swain
wrote: If you check through this thread, Andreas of Novell NZ is kick starting it.
Mike
---------------------------------------------------------------------
To unsubscribe, e-mail: opensuse-unsubscribe@opensuse.org For additional commands, e-mail: opensuse-help@opensuse.org
____________________________________________________ Start your day with Yahoo! - make it your home page http://www.yahoo.com/r/hs
--------------------------------------------------------------------- To unsubscribe, e-mail: opensuse-unsubscribe@opensuse.org For additional commands, e-mail: opensuse-help@opensuse.org
But I have a basic question:
If you are able to do these optimizations, then why not integrate
them
into the main software itself? Why make it a separate downloadable package? Let SUSE *by default* run at high performance. What's the downside to that?
Because it is better to do things gradually and optimise things agressively outside the normal distro and then have those features that prove itself suggested to be put into the distro. I would think that a per CPU optimisation/Prelinking/Kernel patches/rework of the installer is maybe something that is too radical for the mainstream distro right now. On the other side I would be all for it to be part of the normal distro. But I can understand that a more conservative approach is probably better in a distro that must retain the largest amount of compatibility out there.. Would it not be great to have a DVD-ROM with i586 packages for compatibility, but the installer recognize that you run a P4 and install 50-100 specially compiled pentium4 optimised rpm's, just like it currently does for i686? I think this would rock. Then have prelinking turned on by default and use a special desktop kernel in case the user runs the system as a desktop, again optimised for Pentium4. And additionally to that integrate any other technology early. This would be how I imagine a distro to be. Since openSUSE allows me now to create something like this, I want to at least give it a go. You would basically gain the positive effects of LFS or Gentoo or Yoper, but you still retain compatibility with a distro that has thousands of packages available within the community? Well since I think this would be great I am willing to give it a go :) and of course I would love this at some point to be part of the mainstream distro. It is my understanding that the whole reason of openSUSE is to allow new ideas to be integrated easily and in case they are proven to be good ideas possibly integrated into the mainstream SUSE Linux. I love this model and it is exactly such a model which will make openSUSE fly very very long and high ..... Andreas
--- Andreas Girardet
But I have a basic question:
If you are able to do these optimizations, then why not integrate them into the main software itself? Why make it a separate downloadable package? Let SUSE *by default* run at high performance. What's the downside to that?
Because it is better to do things gradually and optimise things agressively outside the normal distro and then have those features that prove itself suggested to be put into the distro. I would think that a per CPU optimisation/Prelinking/Kernel patches/rework of the installer is maybe something that is too radical for the mainstream distro right now. On the other side I would be all for it to be part of the normal distro. But I can understand that a more conservative approach is probably better in a distro that must retain the largest amount of compatibility out there..
Would it not be great to have a DVD-ROM with i586 packages for compatibility, but the installer recognize that you run a P4 and install 50-100 specially compiled pentium4 optimised rpm's, just like it currently does for i686? I think this would rock. Then have prelinking turned on by default and use a special desktop kernel in case the user runs the system as a desktop, again optimised for Pentium4. And additionally to that integrate any other technology early. This would be how I imagine a distro to be. Since openSUSE allows me now to create something like this, I want to at least give it a go. You would basically gain the positive effects of LFS or Gentoo or Yoper, but you still retain compatibility with a distro that has thousands of packages available within the community? Well since I think this would be great I am willing to give it a go :) and of course I would love this at some point to be part of the mainstream distro.
It is my understanding that the whole reason of openSUSE is to allow new ideas to be integrated easily and in case they are proven to be good ideas possibly integrated into the mainstream SUSE Linux. I love this model and it is exactly such a model which will make openSUSE fly very very long and high .....
Andreas
---------------------------------------------------------------------
To unsubscribe, e-mail: opensuse-unsubscribe@opensuse.org For additional commands, e-mail: opensuse-help@opensuse.org
The special desktop kernel idea came to my mind as well ... would be a great thing . Desktop users do not need 100% efficiency so all the magic low-latency patches & kernel tweaks could be included creating a super smooth experience "Would it not be great to have a DVD-ROM with i586
packages for compatibility, but the installer recognize that you run a P4 and install 50-100 specially compiled pentium4 optimised rpm's, just like it currently does for i686? I think this would rock."
-> I'm not completly sure & MIGHT be wrong .. but optimised & fasts distros like Vectorlinux manage to get their speed mostly (??????????) from having slim packages (with rubbish removed .. not sure what exactly that means) & not from special enhancements for Pentium etc. Slackware is supposedly good speedwise but still (????) 486 -compatible. My 26.71 cents of comment . ___________________________________________________________ To help you stay safe and secure online, we've developed the all new Yahoo! Security Centre. http://uk.security.yahoo.com
- > I'm not completly sure & MIGHT be wrong .. but optimised & fasts distros like Vectorlinux manage to get their speed mostly (??????????) from having slim packages (with rubbish removed .. not sure what exactly that means) & not from special enhancements for Pentium etc.
Slackware is supposedly good speedwise but still (????) 486 - compatible.
Yes you are right, that is one approach to get rid of dependencies and it certainly has a huge effect, but you loose functionality, which in a mainstream distro is not good. But with prelinking the library load issue, that happens when you have many dependencies, is mostly diminished as far as I can see (someone correct me if I am wrong) and we really need and want to stay compatible. IMHO I certainly don't want to replace other Linux distro's with this effort, just make sure that the SUSE install I run on my system is as close as possible in speed to those and is maintained even with its customizations. Regards Andreas
On Tuesday 16 August 2005 01:05, Andreas Girardet wrote:
Would it not be great to have a DVD-ROM with i586 packages for compatibility, but the installer recognize that you run a P4 and install 50-100 specially compiled pentium4 optimised rpm's, just like it currently does for i686? I think this would rock. Then have prelinking turned on by default and use a special desktop kernel in case the user runs the system as a desktop, again optimised for Pentium4. And additionally to that integrate any other technology early. This would be how I imagine a distro to be. Since openSUSE allows me now to create something like this, I want to at least give it a go. You would basically gain the positive effects of LFS or Gentoo or Yoper, but you still retain compatibility with a distro that has thousands of packages available within the community? Well since I think this would be great I am willing to give it a go :) and of course I would love this at some point to be part of the mainstream distro.
It sounds interesting, and I'd love to see the numbers, but my gut feeling is that just like with LFS and gentoo, the end result will be less impressive. While I expect there will be gains, I seriously doubt they will be large enough to warrant the effort involved. I suspect the big improvements will come with some serious infrastructure work. X currently isn't doing a very good job, but I know there are several very interesting projects that seem to promise greatly improved performance in the not too distant future. KDE/qt aren't very well optimised, although I hear good things about qt4. gcc isn't the best at optimisations Look for example at the kernel efforts. The major leaps in performance came with projects like the Ordo 1 scheduler and others that completely changed the algorithms used. Recompiling the kernel to use i686 instructions instead of i386 is insignificant in comparison. It's the 80/20 rule in action But like I said, I'd LOVE to be proved wrong, and if I can help out in any way I will, time and work permitting Does anyone have good profiling tools for analysing just where the major bottle necks are? Perhaps studies have been made already? URLs?
Anders Johansson wrote:
On Tuesday 16 August 2005 01:05, Andreas Girardet wrote:
Would it not be great to have a DVD-ROM with i586 packages for compatibility, but the installer recognize that you run a P4 and install 50-100 specially compiled pentium4 optimised rpm's, just like it currently does for i686? I think this would rock. Then have prelinking turned on by default and use a special desktop kernel in case the user runs the system as a desktop, again optimised for Pentium4. And additionally to that integrate any other technology early. This would be how I imagine a distro to be. Since openSUSE allows me now to create something like this, I want to at least give it a go. You would basically gain the positive effects of LFS or Gentoo or Yoper, but you still retain compatibility with a distro that has thousands of packages available within the community? Well since I think this would be great I am willing to give it a go :) and of course I would love this at some point to be part of the mainstream distro.
It sounds interesting, and I'd love to see the numbers, but my gut feeling is that just like with LFS and gentoo, the end result will be less impressive. While I expect there will be gains, I seriously doubt they will be large enough to warrant the effort involved.
I suspect the big improvements will come with some serious infrastructure work. X currently isn't doing a very good job, but I know there are several very interesting projects that seem to promise greatly improved performance in the not too distant future. KDE/qt aren't very well optimised, although I hear good things about qt4. gcc isn't the best at optimisations
Look for example at the kernel efforts. The major leaps in performance came with projects like the Ordo 1 scheduler and others that completely changed the algorithms used. Recompiling the kernel to use i686 instructions instead of i386 is insignificant in comparison. It's the 80/20 rule in action
But like I said, I'd LOVE to be proved wrong, and if I can help out in any way I will, time and work permitting
Does anyone have good profiling tools for analysing just where the major bottle necks are? Perhaps studies have been made already? URLs?
Its not *just* compiler optimisations. There can be large gains in kernel performance based on certain patch sets, -ck is a good example. Certain libs/apps do gain from being optimised, FC do this already to a number of core libs. Compile time optimisation is only one facet of performance. Mike
On Tuesday 16 August 2005 01:31, Michael Honeyfield wrote:
Its not *just* compiler optimisations. There can be large gains in kernel performance based on certain patch sets, -ck is a good example.
Certain libs/apps do gain from being optimised, FC do this already to a number of core libs.
Compile time optimisation is only one facet of performance.
I thought that was what I said So, where do I sign up? And does anyone have any good profiling info to start out with?
On Tue, Aug 16, 2005 at 11:35 am, in message <200508160135.38767.andjoh@rydsbo.net>, andjoh@rydsbo.net wrote: On Tuesday 16 August 2005 01:31, Michael Honeyfield wrote: Its not *just* compiler optimisations. There can be large gains in kernel performance based on certain patch sets, - ck is a good example.
Certain libs/apps do gain from being optimised, FC do this already to a number of core libs.
Compile time optimisation is only one facet of performance.
I thought that was what I said
So, where do I sign up? And does anyone have any good profiling info to start out with?
You just tell me. I can add you to the SUPER website if you want to disclose your name to the bublic and hopefully we will have our own mailing list on opensuse.org soon. http://www.opensuse.org/index.php/SUPER Andreas
Forgive my ignorance: When we say create a SUPER SuSE are we talking about only optimizing the kernel and not the gui apps? Is the point to make a lightning quick server, or a speedy desktop environment? If the answer is both wont we need some help from the likes of KDE, X.org, GNome etc.. ____________________________________________________ Start your day with Yahoo! - make it your home page http://www.yahoo.com/r/hs
On Wednesday 17 August 2005 06:06, Carlos wrote:
Forgive my ignorance:
When we say create a SUPER SuSE are we talking about only optimizing the kernel and not the gui apps? Is the point to make a lightning quick server, or a speedy desktop environment? If the answer is both wont we need some help from the likes of KDE, X.org,
KDE specific issues can get discussed on kde-optimize mailing list. The archive does also cover lots of usual standard suggestions: http://lists.kde.org/?l=kde-optimize&r=1&w=2 bye adrian -- Adrian Schroeter SuSE AG, Maxfeldstr. 5, 90409 Nuernberg, Germany email: adrian@suse.de
On Wed, Aug 17, 2005 at 4:06 pm, in message <20050817040610.57782.qmail@web32510.mail.mud.yahoo.com>, dubbsix@yahoo.com wrote: Forgive my ignorance:
When we say create a SUPER SuSE are we talking about only optimizing the kernel and not the gui apps? Is the point to make a lightning quick server, or a speedy desktop environment? If the answer is both wont we need some help from the likes of KDE, X.org, GNome etc..
Yes and Yes. The more people we will be (especially the programming types) the more we can actively approach those projects and get help there. Andreas
When we say create a SUPER SuSE are we talking about only optimizing the kernel and not the gui apps? Is the point to make a lightning quick server, or a speedy desktop environment? If the answer is both wont we need some help from the likes of KDE, X.org, GNome etc..
Yes and Yes. The more people we will be (especially the programming types) the more we can actively approach those projects and get help there.
Well, I am subscribed to several Gnome mailing lists, being a Gnome user myself. As soon as my new laptop arrives (should be around friday) I can finally install openSUSE. I'll see what Gnome optimisations I can find and report back. Lode
But like I said, I'd LOVE to be proved wrong, and if I can help out in any way I will, time and work permitting
hehehe .. I would love to be proven wrong. But I would also love to be proven right :D That's what is exciting about any intellectual challenge. Like playing chess. Sometimes you win, sometimes you are bound to be defeated. Using the KDrive server might also be ways to get things faster. I would love to try it out and see how good the vesa support is and how fast it compares to your normal X. Have it included and tested and if good available for those who like such bleeding edge in a maintained manner. Another area is early beta releases of kde4 and an initng rpm that converts your existing SUSE install. But booting is honestly already very fast in the SUSE 10 beta1 I am happily running on my laptop, but if we can get it down another 50% I would love to. All those things should be implemented out of the box in experimental releases for those who like that sort of stuff. But in the end this effort will be shaped by anyone who would like to get the maximum out of their installation and be part of it. Overall the first steps of the effort will be to just get our distro into another gear with simple out of the box mods that I suggested before and that are now online on http://www.opensuse.org/index.php/SUPER We will add all the participants in the next days and start organising an irc session. Andreas
participants (9)
-
Adrian Schroeter
-
Anders Johansson
-
Andreas Girardet
-
Carlos
-
Ken Swain
-
Lode Vermeiren
-
Michael Honeyfield
-
Shriramana Sharma
-
Winston Graeme