Anyone on the list using the 2.6 kernel? if so what's the general opinion of it?
Keith
On Mon, Oct 27, 2003 at 05:14:50PM +0000, Keith Watson wrote:
Anyone on the list using the 2.6 kernel? if so what's the general opinion of it?
I've been following the -test kernels on my laptop (running Debian testing). The main thing I'm interested in it for is better power management; it's only since -test7 that suspend-to-disk has started working for me. I can't say aside from that I've noticed anything wonderfully better, but then the laptop doesn't get stressed that much.
J.
On Mon, 27 Oct 2003 17:27:03 +0000 Jonathan McDowell noodles@earth.li wrote:
On Mon, Oct 27, 2003 at 05:14:50PM +0000, Keith Watson wrote:
Anyone on the list using the 2.6 kernel? if so what's the general opinion of it?
I've been following the -test kernels on my laptop (running Debian testing). The main thing I'm interested in it for is better power management; it's only since -test7 that suspend-to-disk has started working for me. I can't say aside from that I've noticed anything wonderfully better, but then the laptop doesn't get stressed that much.
Finally got round to downloading the 2.6.0.test9 Debian kernel source packages.
Everything went like a dream and compiled just like it was supposed to.
The only 2 problems I had were; 1) the proprietary Nvidia drivers need a separate patch applied before you can use them and 2) the Lucent Linmodem drivers aren't available for any kernel after 2.4.x.
(1) isn't really a problem as I found the XFree86 4 NV driver was just as good (for my purposes) as the proprietary ones. (2) was a showstopper for me as I haven't yet got round to buying a 'proper' external modem.
Actually I'm dithering with the modem as I'm also dithering over getting broadband, in which case I'd want at least an ADSL modem.
So it looks like 2.4.20+ for the time being (suits you sir!) :o)
Keith
On Mon, 2003-10-27 at 17:14, Keith Watson wrote:
Anyone on the list using the 2.6 kernel? if so what's the general opinion of it?
I've been using 2.6.0-test9 in a VMware virtual machine for the past few days. It's got Debian Unstable running it it (exactly the same version as running on the host machine, as we have a local mirror). I'm impressed. Mozilla, for example, was quicker at loading, rendering, scrolling and such in the VM than the same version of Mozilla running on the host.
I'll most likely wait until 2.6.2 is released before switching my main machine over to it, though - experience tells me they get a flurry of bug fixes in after .0 and .1 :)
B.
On Monday, Oct 27, 2003, at 17:29 Europe/London, Rob Kendrick wrote:
On Mon, 2003-10-27 at 17:14, Keith Watson wrote:
Anyone on the list using the 2.6 kernel? if so what's the general opinion of it?
I've been using 2.6.0-test9 in a VMware virtual machine for the past few days. It's got Debian Unstable running it it (exactly the same version as running on the host machine, as we have a local mirror). I'm impressed. Mozilla, for example, was quicker at loading, rendering, scrolling and such in the VM than the same version of Mozilla running on the host.
That's the preempt chaps =)
I'll most likely wait until 2.6.2 is released before switching my main machine over to it, though - experience tells me they get a flurry of bug fixes in after .0 and .1 :)
Heh.. I have been running their 2.5.x kernels all the way up to the latest 2.6 kernel. You will definitely love it. Along with andrew morton's patchset, it seriously flies on my mini itx VIA Ezra CentaurHauls 800mhz! Proving this along using mplayer, file sharing (samba/NFS), downloading. Perfect.
I'm pondering to stick it on one of the boxes at work to see if it works in 'real world' situation.
--
- Craig - http://www.wizball.co.uk "Simplicity, the best way to approach life"
On Mon, 2003-10-27 at 18:20, Craig wrote:
On Monday, Oct 27, 2003, at 17:29 Europe/London, Rob Kendrick wrote:
On Mon, 2003-10-27 at 17:14, Keith Watson wrote:
Anyone on the list using the 2.6 kernel? if so what's the general opinion of it?
I've been using 2.6.0-test9 in a VMware virtual machine for the past few days. It's got Debian Unstable running it it (exactly the same version as running on the host machine, as we have a local mirror). I'm impressed. Mozilla, for example, was quicker at loading, rendering, scrolling and such in the VM than the same version of Mozilla running on the host.
That's the preempt chaps =)
Well, I've been using the kernel pre-empt patch for 2.4 for a while. I'm imagining it's actually the much better scheduling and IO systems that do it most. :)
On Monday, Oct 27, 2003, at 19:06 Europe/London, Rob Kendrick wrote:
That's the preempt chaps =)
Well, I've been using the kernel pre-empt patch for 2.4 for a while. I'm imagining it's actually the much better scheduling and IO systems that do it most. :)
Pre-empt, lock breaking and oodles of other little goodies were fun. Since using 2.6 and reverting back to 2.4. I've noticed huge amount of differences between the twos.
Anyone using pre-empt or other patches on servers?
C
On Mon, 27 Oct 2003 20:33:10 +0000 Craig c@wizball.co.uk wrote:
On Monday, Oct 27, 2003, at 19:06 Europe/London, Rob Kendrick wrote:
That's the preempt chaps =)
Well, I've been using the kernel pre-empt patch for 2.4 for a while. I'm imagining it's actually the much better scheduling and IO systems that do it most. :)
Pre-empt, lock breaking and oodles of other little goodies were fun. Since using 2.6 and reverting back to 2.4. I've noticed huge amount of differences between the twos.
Anyone using pre-empt or other patches on servers?
C
I have spoken to some people about this. I have a good friend Kevin who is interested in music, and he has told me about some of his experiments with alsa, low latencies and preemptive kernel patches made, it was also interesting how that they worked together.
My mane interest was to be like windows and be able to play an mp3 some thing I never had done before, under linux this used to just stutter making the idea of playing them too unpleasant to bother with.
me how much difference to audio this sort of patch made on a poi 266 and how my p1MMX 200 would benefit too. Kevin did the full test and is now using both patches as their effect only complements one and other. The low latency and the preemptive after I did it for my own interests.
When I spoke to System administrators (whose systems do calculation and IO) they where interested until they realized that the performance difference was mainly in interrupts and actually reduced server throughput marginally.
Regards
Owen
On Thursday, Oct 30, 2003, at 17:36 Europe/London, owen.churchcowley wrote:
I have spoken to some people about this. I have a good friend Kevin who is interested in music, and he has told me about some of his experiments with alsa, low latencies and preemptive kernel patches made, it was also interesting how that they worked together.
My mane interest was to be like windows and be able to play an mp3 some thing I never had done before, under linux this used to just stutter making the idea of playing them too unpleasant to bother with.
If you look to the other camp.. Mac OS X. They use a mach kernel and it handles a lot of work at the same time. iTunes was chugging away while I load umpteen applications, etc. Anything I throw at it. It just does the job.
Speaking about mp3s in Linux.. isn't xmms is separate from X libs or something? I've read this somewhere *sticks it in the to-do list for research*
When I spoke to System administrators (whose systems do calculation and IO) they where interested until they realized that the performance difference was mainly in interrupts and actually reduced server throughput marginally.
I am currently running this patch on the desktop linux boxes at work:
Con Kolivas - http://members.optusnet.com.au/ckolivas/kernel/
And with some testing (also at home as well before on my mini itx) and it was impressive. However, since I have been playing around with the early 2.5.56+ to 2.6. The differences is really noticable. Mplayer outputted to TV was behaving which was just what I needed along with the use of ReiserFS and a decent hdd (Maxtor 80gig 7200rpm 8mb cache).
Anyway, back on the linux boxes at work. I have spoken to a few people who were victims to the kernel upgrades (I pick them nicely ;). They all agreed that they felt the system was more responsive. This really gives me a little buzz knowing people are going to squeeze a couple of more juices out of those machines!
Blowing my trumpet here about Gentoo.. have you seen their kernel sources list?!
http://www.gentoo.org/dyn/pkgs/sys-kernel/index.xml
Okay, some of them may be a little old but hey, you get choices which is a fine thing.
New subject would be brewing.. what filesystems do you run on your servers/desktop? Reiser4 looks nice so far (But I wouldn't dare run it yet).
And what benchmark tools do you recommend for testing hdds, graphics, kernel load etc? Silly point but would this a bit useful to add to the meetings? Like persuading people to use ext2 ;)
Okay, I'm going to bed!
C
On Fri, Oct 31, 2003 at 12:16:02AM +0000, Craig wrote:
On Thursday, Oct 30, 2003, at 17:36 Europe/London, owen.churchcowley wrote:
I have spoken to some people about this. I have a good friend Kevin who is interested in music, and he has told me about some of his experiments with alsa, low latencies and preemptive kernel patches made, it was also interesting how that they worked together.
My mane interest was to be like windows and be able to play an mp3 some thing I never had done before, under linux this used to just stutter making the idea of playing them too unpleasant to bother with.
If you look to the other camp.. Mac OS X. They use a mach kernel and it handles a lot of work at the same time. iTunes was chugging away while I load umpteen applications, etc. Anything I throw at it. It just does the job.
I think what Owen was trying to say was that playing mp3s on a pentium 233 was very sucky, and I am not surprised to be honest :) I was wondering if he had dma enabled on his disks though, as that can be a real performance killer....
What he says about windows is quite funny though as running linux xmms (or whatever) perhaps uses up to 1% of cpu on my Athlon XP 1700+ but in windows can use up to 80% of cpu and then start to stutter when multitasking...
Although I can make xmms stutter on this machine by disabling dma on the disk and then doing some io related tasks. (well on the old disk, have not tried on the newer disk, but i believe that it would be the same)
Adam
On Fri, 31 Oct 2003 00:38:25 +0000 abower@thebowery.co.uk wrote:
On Fri, Oct 31, 2003 at 12:16:02AM +0000, Craig wrote:
On Thursday, Oct 30, 2003, at 17:36 Europe/London, owen.churchcowley wrote:
I have spoken to some people about this. I have a good friend Kevin who is interested in music, and he has told me about some of his experiments with alsa, low latencies and preemptive kernel patches made, it was also interesting how that they worked together.
My mane interest was to be like windows and be able to play an mp3 some thing I never had done before, under linux this used to just stutter making the idea of playing them too unpleasant to bother with.
If you look to the other camp.. Mac OS X. They use a mach kernel and it handles a lot of work at the same time. iTunes was chugging away while I load umpteen applications, etc. Anything I throw at it. It just does the job.
I think what Owen was trying to say was that playing mp3s on a pentium 233 was very sucky, and I am not surprised to be honest :) I was wondering if he had dma enabled on his disks though, as that can be a real performance killer....
I dont think I had at the time, I have now though, and yes it does make a BIG difference.
What he says about windows is quite funny though as running linux xmms (or whatever) perhaps uses up to 1% of cpu on my Athlon XP 1700+ but in windows can use up to 80% of cpu and then start to stutter when multitasking...
Yes but I was comparing XMMS with winamp which ran on my P200MMX just fine while xmms, and mpg123 used to make a terrible noise when ever I started applications, particularly Netscape.
Although I can make xmms stutter on this machine by disabling dma on the disk and then doing some io related tasks. (well on the old disk, have not tried on the newer disk, but i believe that it would be the same)
You will also notice that X11 sorts its self out quicker (when you switch into and out of it) with the low latency and preemptive patches,
Other things will be much the same.
Regards
Owen
PS sorry for the Typo's in the previous post
On 2003-10-31 00:16:02 +0000 Craig c@wizball.co.uk wrote:
Speaking about mp3s in Linux.. isn't xmms is separate from X libs or something? I've read this somewhere *sticks it in the to-do list for research*
I think xmms can be built without X, yes, but then why not use mpg321 or another console player.
behaving which was just what I needed along with the use of ReiserFS and a decent hdd (Maxtor 80gig 7200rpm 8mb cache).
Is Reiser safe to use now? Here, I've had some fun corruption heisenbugs with a /var on ReiserFS. I think I remember the same problem on the babylonian systems at your work before I left ;-) . I'm now running ext3 in most places. It's not brilliant, but it is fairly well fixed.
There was also some question mark about whether ReiserFS tools are free software, thanks to an unusual interpretation of the GNU GPL by their author. I'm sure searching for Reiser on debian-legal will return many messages. http://lists.debian.org/
On Fri, 2003-10-31 at 08:09, MJ Ray wrote:
Is Reiser safe to use now? Here, I've had some fun corruption heisenbugs with a /var on ReiserFS. I think I remember the same problem on the babylonian systems at your work before I left ;-) . I'm now running ext3 in most places. It's not brilliant, but it is fairly well fixed.
I have been running ReiserFS safely on a production filesystem on the Pepperfish servers for over three years now. The only thing tempting me away from it is XFS ;-)
D.
On Fri, 2003-10-31 at 00:16, Craig wrote:
If you look to the other camp.. Mac OS X. They use a mach kernel and it
Mmmm mach -- the only instance where I can point and say "haha! your microkernel is bigger than my monolithic one" (c.f. the Ely meet on Wednesday)
The only difference between Linux and Mach from the point of view of low-latency and kernel preemption is that mach was written to be reentrant from the start, and the Linux kernel has had to have it hacked on the side.
Linux 2.6 is much much better than Mach, no doubts about it.
D.
-
abower＠thebowery.co.uk -
Craig -
Daniel Silverstone -
Jonathan McDowell -
Keith Watson -
MJ Ray -
owen.churchcowley -
Rob Kendrick -
Rob Kendrick