“For high-end and server applications, probably the most significant [upgrade] is in scalability,” says Andrew Morton, who — along with Linus Torvalds — maintains the Linux kernel. “2.6 scales to more CPUs, more memory, more disks and larger disks,” he told NewsFactor. Here is also a discussion about Linux kernel 2.7.
Review: Linux Kernel 2.6
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
It would be nice to read the article; if it was /. of course.
Is 2.6 everything that the zealots have been screaming that 2.4 already is (the most stable, responsive, easy to use desktop operating system EVAR)?
it is, infact, it is even more than that trolls give it credit for.
Don’t start a flame war. LINUX is used it many mission critical places (you are probably using it now), so its obviously stable and responsive.
2.6 is (SHOCK) better than 2.4.
How long has MS been touting how secure Windows is(in between Virus reports)?
No its not, its LINUX(the kernel).
We’re not talking about a distribution.
If that was a joke, its not funny, how would you like to devote 20 years of free work, and not get any credit?
If that was a joke, its not funny, how would you like to devote 20 years of free work, and not get any credit?
Maybe he should have used a license that requires credit in the source files, like the BSD. I’m sorry he wasted 20 years of his life because of his own stupidity, but I’m not going to cry for his mistakes.
He used the license he wrote. Why would RMS write a license, and then not use it?
And read the BSDL again, it doesn’t require that.
It doesnt even require SOURCE to be distributed.
He used the license he wrote. Why would RMS write a license, and then not use it?
How about, why would I feel sorry for him not getting credit, when he did not require so in the license he wrote?
And read the BSDL again, it doesn’t require that.
I said BSD requires credit in the source files.
http://www.opensource.org/licenses/bsd-license.php
Read it.
It doesnt even require SOURCE to be distributed.
OH WOW! Who was arguing that? Oh yeah, no one.
Hey, who cares about credit in the source files? I bet 99% of Windows users don’t know and will never know that Windows contains code from BSD. Anyway, it’s not like anyone who reads source doesn’t know where all the basic userland tools in GNU/*/Linux come from.
I certainly don’t, but “contrasutra” (if that’s your real name) is crying a river:
how would you like to devote 20 years of free work, and not get any credit?
“Is 2.6 everything that the zealots have been screaming that 2.4 already is (the most stable, responsive, easy to use desktop operating system EVAR)?”
Yes, only it’s done up in a fashion even a lumox could find useful. Give it a try.
No, because we’re talking about a kernel, not an OS.
Oh of course, leave it to the geeks to get all technical. Of course no one was technical and mentioned the fine print when they were telling me Linux (or is it LINUX) is the most stablest, rock solidest operating system ever. They never mentioned that this is because there are no programs for it that do anything even as a complicated as cut & paste.
Just because I’m not giving out personal information doesn’t mean you can’t have a serious conversation. Well, maybe it does, but I doubt they are correlated.
Oh of course, leave it to the geeks to get all technical.
I don’t think you understand. Usually, I steer away from the “Linux vs. GNU/Linux” controversy. I call the OS Linux (or Mandrake Linux, or RedHat Linux, etc.). However, in this case, they’re really talking about the kernel, not a distribution. It’s not a matter of semantics, but of knowing what you’re talking about. Since (obviously) you don’t know what you’re talking about, I though I might take the (futile) opportunity to enlighten you.
They never mentioned that this is because there are no programs for it that do anything even as a complicated as cut & paste.
Heh. You’re funny. Gee, if there’s no cut and paste, I wonder how I reprinted your inane drivel…
Of course, there are tons of apps for Linux, and you can also run the most popular Windows programs on it. But trolls will be trolls, I guess.
Hey, moderators…how come you let such garbage pass? Do I detect a bias here?
You are not a troll because you dislike Linux. You are a troll because you deliberately make inflammatory (not to mention false and off-topic) comments with the clear aim of provoking a strong reaction among others. That is the definition of trolling. Here’s one of many interpretations:
http://c2.com/cgi/wiki?TrollDefinition
For some reason, your trolling has not been modded down. Personnally, I don’t mind – it supports my opinion that there are as many anti-Linux trolls as there are Linux zealots trolling on MS threads.
While you seem to be unable to refute my argument,
Argument? Where? I must have missed it. Let me see, now..
Hm. Are you sure you had one? Because I’ve re-read your posts and I sure can’t find one. Some false statements, some FUD and lots of trolling, but not a single argument. Perhaps you would like to formulate it now, so I can shoot it down properly?
you have managed to quote a slogan beloved around the world by 10 year old girls.
I was trying to speak in words you understand, so I stooped down to your level. But perhaps I didn’t stoop low enough…
Actually, Ive changed my mind. Perhaps I should use a username and not go as just Anonymous. After all, how else can you even know its me?
😉
Appologies to the REAL anonymous in this conversation for impersonating you, but I believe I made my point. After all, I have NEVER registered or givin out personal information, and yet I do not go as Anonymous.
Posted to soon then. It appears Anonytroll has registered.
He’s really putting up an excellent fight ;o)
Well done, anonymous, show the Linux zealots!
Linux: decades in the development, and still catching up.
What I’d really like to see are some tests that compare applications being run to completion on the new kernel and the 2.4 kernel whilst under load. It’s all very well going on about responsiveness, but how much overhead is that extra responsiveness causing?
I mean it seems pretty obvious to me that if in one case an mp3 skips whilst doing heavy compilation in the background and in the other the mp3 doesn’t, then when the mp3 isn’t skipping the compilation is getting less cpu time. Since the song isn’t being played any faster how much longer does the compilation take as a result of the reduced processor time? I’d like to see a few comparisons along those lines instead of things like “It takes xxx nanoseconds less to respond when I hit the mouse”.
I guess there’s a bigger issue here. Namely how much power people are prepared to trade for responsiveness. If you make the GUI responsive as possible can a kernel get away with running background tasks at very low priority without compromising the power of the OS?
The other question would be “Do the Linux developers have eyes bigger than their stomachs?”, because it really does seem sometimes that they’re trying to create a one-size-fits-all operating system. I respect their work and everything, but occasionally I get the impression there is a lack of genuine focus among the development group that leads to people pulling in different directions (Which I admit isn’t always a bad thing).
Can you explain?
Is this a reference to the common misperception that most open source developers are starving, and subsist soley off of Paypal donations?
(Please note the word ‘misperception’ before you are offended)
And do you dare deny that Linux is not a community founded on slander?
Why, yes, I do deny it.
The Linux community fails to get ahead by making their own product better
Actually, it is getting ahead, and the product is getting better.
(some kind of problem with hiring developers to work for free)
Oh, haven’t you heard? A lot of people contributing to Linux development are actually hired by some of the largest IT companies and most prestigious institutions: IBM, HP, MIT…
so they instead sling mud at other companies.
No, that would be SCO.
They try to spread “fear, uncertainty, and doubt”
No, that would be you.
Since you seem to be unable to do any research,
Sure I’m able to. But I don’t feel like wasting more time than necessary on trolls.
I’ll provide two links for you to read (at your convenience of course).
I read them when they came out. The ESR piece is actually tame when you compare it to what Wall Street companies had to say about Sun. The writing’s on the wall, ESR just pointed to it. No FUD here.
The Roblimo piece is a tongue-in-cheek look at Windows from someone who’s never used it. It’s written in a light tone but actually makes a few valid point. No FUD here.
Meanwhile, I’d like to destroy another one of your faulty deductions. You wrote:
since there are billions of Windows users, and only dozens (if not hundreds) of Linux users – and the number of trolls comes out about even from both camps – I’d say a much larger percentage of the Linux camp is “trollering.”
Actually, according to the latest OSNews OS poll, there is in fact more Linux users on this site than Windows users. Since there is an equal amount of trolls on this site from both camps, it would seem – following your logic – that there’s in fact a larger percentage of the Windows camp that is trolling.
http://www.osnews.com/story.php?news_id=3775
Linux: decades in the development, and still catching up.
Gee, how many decades has Linux been in development? Mmmh…let me count: one!
I think what you meant to write is:
“Linux: only a decade old, and already surpassing Windows.”
I find it strange that someone from Nokia – which originates from Finland – would poo-poo on Linux, the creation of a Finn.
So, what about those exploding phones, eh?
Well, I’d love to hang out and talk some more, but I’ve promised my gf I’d spend some quality time with her, and stop feeding the trolls. Sorry, guys!
Just ignore these trolls,they’re not even interesting.
Their parents will come home soon enough and they’ll shut up.
Anyone know what happened to topspeed and that snakeguy?
At least I responded to your post, instead of trying a diversionary tactic to cover the fact that you don’t have any counter-arguments.
I believe that is cheating. Goodbye, cheater.
You’re right, I’ve been trolled long enough. He’ll dry up and go away eventually, just like TopSpeed and CoralSnake. Too bad there’s not “ignore” feature, though (and too bad the moderators are sleeping on the job).
Yes, it’s too bad the Linux supporters can’t permanently silence all of their critics. Someday, if you achieve a Linux society, with a Linux-controlled government- then you can silence all of the dissidents, but until that time – you will have to either read Slashdot, or use the scroll button on your browser.
That would be neat if OSNEWS had an “ignore” feature where you could ignore certain IPs (attbi).
Eugenia: Any chance of getting this. I know you dont want a “slashdot” type system, but this sounds reasonable.
Since there were so many OT posts between this and the original ralid question, I’ll repeat…
The original asked the question of “How much application performance is lost due to the improved GUI responsiveness”
I would also like to see a comparison between applications run on 2.4 and 2.6. I suspect that is is fairly small, but I could be wrong.
However, the features that have been added can be turned off at compile time. If you know you are making a kernel for servers, you can disable preempting or you can turn it on for a workstation or home system.
The improvements to the VM and many other improvements help both servers and home users so I would think 2.6 will still be the better kernel overall.
BTW, where are the moderators?
Mutiny
In my previous post, I meant to say VALID post, not RALID and my caps lock was on so I am not mUTINY, but Mutiny…
Edit button PLEASE!!!!
Mutiny
OK. Ill answer your question, keeping in mind im no kernel developer.
I’ve been running the 2.6 series for a while (-mm patchset) and the multitasking really is wonderful. I can do 10 things at once with no “slow down”. Listen to music, surf the web, compile, GIMP it up…
Now, yes, when compiling and doing other things, the compiling DOES slow down, signifigantly. The advantage is that I can keep it running in the background, and even though its running slower, I can keep up normal work, because the UI stays responsive.
Now, if Im JUST compiling, that runs much quicker than in 2.4.
Linux may not be a great desktop for the windows “power user”, it is great for the extreme parts of the spectrum, Joe User and the geeks. The problem is that, the “power users” are used to be able to muck around the system in a very windows-like way, so when they come to linux they need to relearn everything. On another note, recently, the GNOME desktop has improved alot, I really enjoy it. The font rendering is top notch (very improved from two years back IMO).
I would easily say that the GUI, to me is better than Windows, but the intergration with the core part of the system does need some work. Kernel 2.6 is great, really great responsiveness and improvements all around. The desktop works really great intergration-wise (such as cut-n-paste) in a homogenous desktop enviroment (gtk and gnome apps only, or qt and kde apps only.) though I cannot speak for kde. I think that the problem with intergration can be solved if distro makers concentrate only on one demographic, instead of being a one size fits all bag. Have a seperate version for Server and Desktop. Right now GNU/Linux is still rough around the edges on the desktop, but with some polish on the part of the distros, it can be great.
On anonymous’s part: I’m sure linux is destroying your life. oh wait it isn’t, shut up. It’s almost as if you fear linux, the way you talk about it.
I haven’t been able to get NVIDIA drivers to install so I’ve decided I’ll behave and wait for the final release of 2.6.0, preferably using Mandrake’s package.
Multitasking is good – with a Gig of RAM I keep my desktop programs open all the time (and still have some room) and I’d enjoy compiling in the background once in a while.
Dammit, I want it NOW! 🙂
>“I’ve been running the 2.6 series for a while (-mm patchset) and the multitasking really is wonderful. I can do 10 things at once with no “slow down”. Listen to music, surf the web, compile, GIMP it up…”
Geee can it make toasts too? 😛
YES 2.6 is faster but I doubt anyone can tell the difference just by using it.
YES 2.6 is faster but I doubt anyone can tell the difference just by using it.
Actually, the consensus seems to be that it “feels” faster, and that the UI is more responsive. Not a dramatic improvement (after all it’s already pretty good), but not so subtle that you don’t notice it. And in UI improvement, every bit is always appreciated.
P.S. with the new version of xscreensaver, the toasters from After Dark make a comeback. So it does make toasts. In a way.
The kernel feels slower here because of some issues with the new scheduler and my computer. Well, that’s what I’ve noticed with -test4. Perhaps I should try -test7 to see if they have fixed that.
Actually, I CAN tell just by using it. I only have a 2Ghz P4, im not running some massive server doing huge calculations, its a desktop machine.
For example, webpages render twice as fast(no kidding) and I can flip through Mozilla tabs instantly.
How is that not “the user feeling it”?
Also, programs “pop up” faster as well.
Not to mention UT2003 running much better.
I ran kernel 2.4.19 on my laptop, a Pentium-3m 1ghz, with 512mb RAM. I had problems with my music skipping under heavy load, and even my mouse not moving smoothly. I did not use any of the interactivity patches though, as I’m more of an “end-user” than anything else. Also, it seemed that Mozilla and Mozilla Firebird took a little longer to start than the equivalant on Windows (I won’t say how long to avoid a flame war). Just curious if anyone thinks the new kernel would speed up application launch times and improve interactivity by a noticable amount to a typical end-user.
>“For example, webpages render twice as fast”
“Pull man! Pull! I think you’ve got me!” : [the sassy mouth salmon after hooking Johnny Bravos fishing hook to Johnny Bravos pants]
Probably, as I understand it the scheduler for 2.4 wasn’t that great, for desktop use anyways. I would try 2.6 with the -mm patchset, it’s pretty awesome with his new scheduler, and the new interactivity patches. Or try the -ck patchset for 2.4 even.
I use Thomas Backlund’s patched kernel and sources for Mandrake, and even though desktop responsiveness is pretty much the same as for the regular kernel, the multimedia is much better, it never skips and runs smooth even if I switch desktop – heck, even if I switch to another X session running at the same time.
I can’t wait for 2.6. Did I mention this already?
I suggest you stop trolling on each other, or you will find this thread closing and ALL comments deleted, PLUS your IP address banned.
Does anyone have a rough idea (+- a few months) when they think some of the bleeding edge distributions will start shipping with the new 2.6 kernel? Just curious.
I think the kernel will be released around December, so distros released after that will most likely start around January-February.
Suse includes the latest 2.6 test-kernel with its new distribution SUSE 9.0 shipping in a few weeks.
>Does anyone have a rough idea (+- a few months) when they think some of the bleeding edge distributions will start shipping with the new 2.6 kernel? Just curious
SUSE are going to ship with a test kernel as an alternative next month in v9.0
http://osnews.com/story.php?news_id=4680
Of course ‘bleeding edge’ distros such as Gentoo can probably emerge it already. All others will no doubt have packages too. There is always compiling it yourself!
“Now, yes, when compiling and doing other things, the compiling DOES slow down, signifigantly. The advantage is that I can keep it running in the background, and even though its running slower, I can keep up normal work, because the UI stays responsive.
Now, if Im JUST compiling, that runs much quicker than in 2.4.”
I’m no linuxphile or linuxphobe but licurious
So a way around this would be to compile two kernels and have them both as boot options to reboot depending on what you were wanting to do? Or am I overstating the simplicity?
I find it strange that someone from Nokia – which originates from Finland – would poo-poo on Linux, the creation of a Finn.
Strange? Not at all, if you consider how overhyped Linux is. For your information, I started to use Linux in 1996, when I was just a humble network project manager. I started with Slackware, and liked it. A lot.
However, after all these years, I feel that Linux has not lived up to the repeated promises of it’s developers and community. And even worse: I have worked a lot with HP-UX and Solaris in the meantime, so I learned how a good UNIX kernel should work internally. I also started using BeOS (I “met” with BeOS r. 4 in 1997, IIRC) and was completely blown away. What an OS!
So I had several multibooting computers, at work and at home. Most of them had both Linux and BeOS (in addition to some other, more arcane OSes). I could compare the behaviour of Linux and BeOS; Linux always, invariably, left a lot to be desired. What a shame, that after all this time, BeOS still smokes Linux for responsiveness and autoconfiguration.
And what a shame that after all this hype and money thrown at it by IBM and HP, Linux is still behind unices like Solaris or Irix, in throughput and thread management.
And the price? RedHat AS Linux, the market leader (in corporate Linux server installations) and the only Linux supported by most 3rd party SW vendors, costs an arm and a leg, yearly, PER SERVER! And then you get a really crappy support.
As for Finland vs. the creator of Linux: you would be surprised to know how many senior IT professionals are annoyed by the HYPE around Linux. Gradually, a very strong opposition is emerging. More and more benchmarks are made, and the king seems more and more naked by the day. No wonder that not even IBM customers, who use AIX, are convinced.
wasn’t turbolinux already shipping a 2.6 kernel?
and suse is providing it as an option i guess
There are already some distros distribuing with the 2.6 test kernel (Turbolinux comes to mind). Debian Unstable and Gentoo also have 2.6 kernel packages. Gentoo are also including patches for packages that doesn’t support 2.6.0-tests. I suppose they’ll offer 2.6.0 as soon as it’s out. It’ll probably still use devfs, though.
“”Just curious if anyone thinks the new kernel would speed up application launch times and improve interactivity by a noticable amount to a typical end-user.””
Can’t comment on the interactivity improvement (Although reports seem to show it is definitely noticeable). However launch times (First time launch, so no caching involved) are unlikely to improve much at all since the largest bottleneck is due to the low speed (In comparison to the CPU) of disk operations which has almost nothing to do with the kernel (Although there are both efficient and inefficient methods of controlling disk access).
(second attempt to post this… first went through but never appeared?)
“Now, yes, when compiling and doing other things, the compiling DOES slow down, signifigantly. The advantage is that I can keep it running in the background, and even though its running slower, I can keep up normal work, because the UI stays responsive.
Now, if Im JUST compiling, that runs much quicker than in 2.4.”
I’m no linuxphile or linuxphobe but licurious
So a way around this would be to compile two kernels and have them both as boot options to reboot depending on what you were wanting to do? Or am I overstating the simplicity?
The apps in linux don’t launch slowly because of the HDD. It is because of all the libraries that get launch for that one app.
I think you can do something called prelinking (to prelink the libraries) so it can launch faster.
Mandrake has prelinking system wide as seen on a recent OSNEWS article. But frankly I saw no difference in launch times to any other distro out there.
I wish something could be done about this because win2000 runs beautifully on my k6-2 450Mhz and apps like Mozilla launch way faster and there is no delay in the interface. The 2.6 kernel may help with the latter.
I guess instead of close programs it is better to move them to a new virtual desktop. But the thing is, that is not how I use my computer, I like to shut it down when I am not using it.
Anyone using devfs or devfsd with 2.6? Thoughts, comments etc would be grateful.
Doug
The apps in linux don’t launch slowly because of the HDD. It is because of all the libraries that get launch for that one app.
I think you can do something called prelinking (to prelink the libraries) so it can launch faster.
Mandrake has prelinking system wide as seen on a recent OSNEWS article. But frankly I saw no difference in launch times to any other distro out there.
I wish something could be done about this because win2000 runs beautifully on my k6-2 450Mhz and apps like Mozilla launch way faster and there is no delay in the interface. The 2.6 kernel may help with the latter.
I guess instead of close programs it is better to move them to a new virtual desktop. But the thing is, that is not how I use my computer, I like to shut it down when I am not using it.
Does anyone have a rough idea (+- a few months) when they think some of the bleeding edge distributions will start shipping with the new 2.6 kernel? Just curious.
Good question. I would say that the first distro will most likely be SuSE followed by either Red Hat or Mandrake, however, that would depend on the state of the Linux kernel once 2.6 has been released.
Many distros were really stung the last time when they jumped on board early only to find that there were major bugs still yet to be wacked out of the kernel, such as IDE filesystem corruption, baroque VM (fixed in 2.4.13 and greater releases).
If it were me, I would wait until 2.6 reached around 2.6.4’ish so that any shockers are bashed out of the kernel before using it as a full time kernel.
One thing I would like to see is an update on how XFS and JFS are now performing on 2.6 vs 2.4. If anyone is running those filesystems and the 2.6 kernel, could you update the “community” of your experience?
I use DevFS with 2.6. Runs fine, I only had an issue with -test5-mm3, but thats long fixed.
I really don’t think most understand the significance of the 2.6 kernel release. This goes a long way in addressing some of the basic problems I had with Linux kernel. The thing I am most excited about is a new user level threading model and a lot of big kernel locks are now fine-grain and finally NUMA. There are numerous improvements, some of which are dramatic (look at devices and kernel modules.) 2.6 is a big deal for Linux. http://www.kniggit.net/wwol26.html
I wish that Solaris and FreeBSD (and more recently DragonFly BSD, take a look at their scheduling, etc) were recognized for their sophistication and advanced structure/algorithms. Look at BSD’s Vitrual Memory model and their threading with KSEs and trying to get the kernel be free from large kernel locks, and certainly the work they have done with jails. Take a look at Solaris’s thread model, how it handles memory and I/O, it has had an O(1) scheduler for years, It has multiple schedulers running for different threads at the same time, and the fact the kernel is COMPLETELY pre-emptible. No, Linux STILL doesn’t have this (although 2.6 goes a long way in helpling that.) I understand that everyone has their “favorite” OS, but these are usually the OS’s that remain innovative. Windows 2000 has also had its fair share of innovations, but BSD and Solaris still remain on top in my book.
“”The apps in linux don’t launch slowly because of the HDD. It is because of all the libraries that get launch for that one app. “”
My comment wasn’t Linux specific. HDD access times ARE the limiting factor on every single OS that uses HDDs to store their applications. No matter how many tricks are pulled with caching, pre-linking, or static compilitation sooner or later the application has to be brought from the HDD and into memory. The bonus to using common shared libraries (Aside from the obvious ones of efficiency) is that over time the system will have the majority of the common libraries held in its caches, removing the necessity to retrieve them from the HDD.
There have actually been some interesting studies done (I’m
not sure if this URL is totally legal. I think the IEEE normally charges for these papers. http://www.cis.upenn.edu/~mwh/papers_DB/ieee_computer97.pdf) on using slim binaries that are compiled at runtime instead of pre-compiled code that is linked at runtime. It’s slightly slower, but only slightly, because the disk access overhead is so large compared to the linking overhead. So by reducing the size of the binary (In this case to their slim binary format) the improvement in HDD overhead very nearly outweighs the benefit of code being precompiled. As processor speeds continue to follow Moore’s law, and memory is becoming insanely cheap, this may be the future.
Well, Solaris and AIX may still perform better than Linux, but still people are switching away from them to Linux-run Intel machines. For the same price, they end up with more processing power and better overall performance. Linux’s horizontal penetration is also unprecendented for a Unix-like OS (remember, it’s not Unix). You don’t see many supercomputers made up of Solaris, HP-UX or AIX nodes, or embedded devices running those (admittedly powerful) Unices…
I feel you are unnecessarily harsh with Linux. The rate of development, though chaotic, is unprecendented, and it has lived up to all of its promises, save for one: supplanting Microsoft, which was never a reasonable expectation in the first place.
Linux may not (yet) be as powerful as Solaris, HP-UX oar AIX, but it’s getting closer every day. However, it has done something that no other Unix had done before: it has democratized the Unix-like experience, and that’s why you’re seeing IBM and HP focus away from their proprietary Unices towards Linux. As for Sun, well, their Linux strategy (or lack thereof) is partly responsible for their current sorry state.
If Linux was such a disappointment, do you think 70% of Wall Street banking firms would be using it?
As for BeOS, I think what happened to it is really sad, but it goes to show that you can’t go head-to-head with the monopoly on its own turf, fighting it with the same weapons. The reason Linux (and, to a lesser extent, the BSDs) are slowly gnawing away at MS is that they are open-source and cannot be bought off or forced into bankruptcy. That, more than any technical superiority, is the real strength of Linux – the technical superiority over the other Unices will come in due time, as development continues.
My 0.02$, anyway.
I actually managed to compile and get running a Linux 2.6-test7 kernel despite being almost a complete noob (I have been using Debian as my first try at Linux for about a month). To me, the better system responsiveness is very noticeable compared to 2.4. I do beleive however that this is something that matters more to some people than others. Personally I find it important that everything about the GUI happen instantly, without any delay. 2.6 feels a LOT more crisp in this regard, and more so than my xp installation on the same machine.
I have heard a lot of people remark that one of the most significant indicator’s of “fast feel” is that their mp3 player doesn’t skip when doing large compiles. I have only noticed that problem when I was running a stock debian 2.2.20 kernel. After I upgraded to 2.4 and set up hdparm for IRQ unmasking (sp?), I have never had a problem since. If I compile something that has a lot of small files (say, GCC-3.4 (experimental) for example) than, yes, the rest of the UI becomes choppy, but XMMS *never* skips. Solely to keep everything else responsive, I can nice up the compile with `nice -n 19 make`. That, IMO, is far better than running X with negative niceness. Finally, if I compile something with lots of large files, than nothing slows down.
Since my system is a rather old 800MHz PIII with only 384MB of RAM, I have to wonder if the folks that complain about MP3 player skipping simply have something screwed up in their system config. Try playing with hdparm (after reading the man page, of course). After backing everything up.
Overall, to me, GNU/Linux with the Linux 2.4 kernel feels *far* more responsive and fluid overall than Windows XP on the same hardware.
That’s interesting.. What hardware is that exactly? I had the opposite experience. Mozilla and Firebird seemed to take more than twice as much to load, in cache and out. I have never had my music skip in Windows, but it did happen occasionally in Linux.
I wonder why no MP3 player (at least I know of none) preloads the whole song and then plays just from memory. Usually it should be less than 10 MB (or less than 5 MB if you stick with 128k or less like me) and then a whole lotta problems would be gone… Of course they had to make sure that the song wouldn’t be paged out to swap immediately…
Philotech
As a Finn I decided to take part with a comment.
I can understand your sentence:
“I find it strange that someone from Nokia – which originates from Finland – would poo-poo on Linux, the creation of a Finn”
But then again Windows is american and I believe there are some americans who would poo-poo on Windows too.
Se the point.
About Marios rant, Mario is probably not a Finn as Mario is not a finnish name, and Nokia is a very very international company.
This rant from Mario:
“As for Finland vs. the creator of Linux: you would be surprised to know how many senior IT professionals are annoyed by the HYPE around Linux. Gradually, a very strong opposition is emerging. More and more benchmarks are made, and the king seems more and more naked by the day. No wonder that not even IBM customers, who use AIX, are convinced”
– is rubbish, but OK it is, perhaps his oppinion and probably about the desktop.
Linux is growing in Finland as in all other parts of the world.
About Nokia. Nokia is using linux in their set-top boxes and
also for their IP stuff. They are members of the OSDL and take part in the carrier-grade stuff.
So far there is no Nokia linux smartphone, as they are dedicated to Sympian, but we will se, if the pressure from
Motorola, Samsung and MS gets heavy then who knows.
Not such a bad idea with a handheld with a real-time multitasking OS.
Regards,
In addition to the CPU and memory I mentioned before, the HDD is a 60 GB, 7200RPM Maxtor with 2MB of HDD buffer. The interface is ATA-100. The command I use to tweek it is:
hdparm -q -c3 -m16 -d1 -u1 /dev/hdb
/dev/hda is for my very idle Windows drive.
Don’t play with hdparm until you read the manual for it. I didn’t have any problems, but the text basically starts off predicting doomsday consequences if you aren’t careful.
With regard to preloading songs, I think that the Linux VM will see that you are using that file continuously, in essentially read-only operation, and give it higher priority to keep it buffered in local RAM. I have a project that takes nearly all of my available memory to compile (for some files), and even when my webrowser and IDE get paged out to disk, XMMS keeps on ticking.
Of course, it’s silly to think that all Finns would automatically admire Linux. After all, I’m from Quebec and I hate Celine Dion (well, I hate her music, anyway.)
Point well taken.
I find it odd that you are not a fan of the “Canadian Dog Sled” since you are from Canada.
http://www.campushook.com/?pg=profile&u=1866
I have played back ten simultaneous tracks of 44.1k/32bit audio in Ardour, with gverb and a couple of alsa eqs and compressors, while compiling Mozilla in the background, and browsing usenet (though knode was running on another machine over ssh).
That was on a PII 400 with 256MB, and ancient 1.4Gb drives, with 2.4.22+Low latency and pre-empt patches.
I had no skips or x-runs in the audio. At a 256 *samples* latency!
My configuration would probably suck for a server, but it’s a nice desktop. So don’t take xmms ‘skip tests’ to be the gospel truth as far as audio on Linux goes, alsaplayer, sweep, Ardour are all much more critical about making sure your audio remains uninterrupted.