I found this interesting and thought I would share.
Linux currently is using CFS (Completely Fair Scheduler)
Con came out with BFS:
BFS was created to improve responsiveness on Linux desktop by providing lower latency. So, you could run multiple applications at same time without them stalling out. BFS could only handle upto 16 logical cores before taking performance hit. CFS works with 4096 logical cores. http://ck.kolivas.org/patches/bfs/bfs-faq.txt
A comparison of CFS versus BFS:
You should search for “BFS vs. CFS Scheduler Comparison. Taylor Groves, Jeff Knockel, Eric Schulte” because I had trouble posting the link.
[quote=tonestone57]Rox wrote:“I’m guessing you are simply someone with a grudge against Linux for some reason and now you are simply trolling.”
I’m the one that said Linux is targeting and being used for desktop OS and I’m trolling? Look again. You’re the one that is trolling saying Linux is not for desktop when even Wikipedia agrees with me and you went on with this big rant to say you were right about this point when you clearly weren’t. Showing you have strong bias and feelings on the topic.
Yes, Linux has taken off on server market, everyone knows that but the distros are for desktop (individual use). Something you fail to see yourself. Who do Ubuntu, Fedora, OpenSuse, Sabayon, PCLinuxOS, Mandriva, etc. target? Desktop users. Yes, these distros can be used for servers too but they were created for the average computer user to run at home.
“Linux (commonly pronounced /ˈlɪnəks/ LIN-əks in American English,[4][5] also pronounced /ˈlɪnʊks/ LIN-ooks[6] in Europe and Canada) refers to the family of Unix-like computer operating systems using the Linux kernel.”
“Typically Linux is packaged in a format known as a Linux distribution for desktop and server use.”
The information I provided about Linux is accurate and true! I have not made it up like you have about not marketing as desktop OS. I even gave links to support my comments where possible and even 2 cases for making of the CFS patch; one of which is right, but maybe reading the links I gave is too hard for you? Linux had responsiveness issues - 2 different situations. Patches were made to fix these issues. I pointed the issues/patches out with links. Maybe you wanted to sweep these issues under the rug and hide them? X has performance issues too. Why else are developers working on Wayland to replace it? I can pretty much prove everything I said so how is that trolling? Just because you disagree with me? I gave true information and supportive links. Maybe you’re not used to technical analysis? These issues were real and recent. I just didn’t make them up for the fun of it.
I am very happy that Linux is now looking at improving and fixing OS responsiveness and focusing equally on desktop use. Sounds like you’re the one upset about this. Haiku will have a tougher time competing with Linux because of changes like these which improve latency and performance.[/quote]
I really don’t see linux competing with anything on the desktop with its less then %1 marketshare on the desktop. If linux is competing with anything on the desktop, its MAC which has a 4-5% market share.
Its great they are fixing the responsiveness issues. How are they going to make it point and click friendly ? Truth is, the linux elitist mentaility will doom the OS to a command line type structure for the forseable future.
I really don’t see haiku having a problem competing with linux, becuase linux isn’t competitive in the desktop space as it sits now. hiaku shouldn’t target linux users, total waste of time.
The BFS vs CFS scheduler comparison is found here:
Newer CFS should run as good or better than BFS now!
I found it very interesting that Con had to give a push (with BFS) to get Linux to improve itself. Why it took so long to make this 224 line change makes me really wonder what is going on with Linux.
ie: Con had to release BFS to get someone to improve CFS scheduler and it took just over a year, from BFS getting out, to get this done.
The recent suggestion is not in any way equivalent to Con’s BFS. Con wanted a different scheduler design, and when Linus didn’t accept his proposal (and thereby make him maintainer of the scheduler rather than Ingo) Con basically ragequit, confirming Linus’ belief that he was not a good choice of maintainer. As often happens when people ragequit, Con is now back working on Linux.
The “200 line patch” puts processes into separate task control groups based on their virtual terminal. This means that when you create an xterm (which uses a virtual terminal) the programs you run in that xterm can be given different scheduler treatment from those running as batch jobs, without the user needing to do anything extra. Linus apparently liked this change, not least because it’s small and easy to understand.
cgroups have existed for a long time in Linux, this patch just provides an intuitive way for them to come into existence and be destroyed without any administration.
Scheduler tweaks are subject to a powerful placebo-like effect. Cyanogen reported that adding BFS to kernels made users report much improved latency - but others have reported the same from removing BFS. Google looked at including BFS in Froyo and concluded that blinded users (ie those who aren’t told which they’re using) can’t tell the difference between then in Android.
My first point is that this “200 line patch” came about to combat latency and improve responsiveness of Linux and improves CFS to compete against BFS. Would this change have happened if BFS did not exist? I do not think so. Con motivated Linux developers to change CFS to work better otherwise CFS would have gone unchanged for more years or forever.
My second point is that CFS was first included in kernel 2.6.23 released around October 2007. It took 3 years to make this small change to improve scheduler performance on Linux. Why didn’t anyone look into doing this sooner?
“The Linux kernel received a patch for CFS in November 2010 for the 2.6.38 kernel that has made the scheduler fairer for use on desktops and workstations. Developed by Mike Galbraith using ideas suggested by Linus Torvalds the patch is expected to significantly boost multi-tasking performance on most systems in that class.”
Yes, this completely unrelated change would likely still have happened without BFS. Obviously I can’t make you change your mind but basically your position is as if you’d asserted that without Truman Capote writing “Breakfast at Tiffany’s” there never would have been TV SF shows like Star Trek. It’s silly, but nothing about it can be proved without rerunning history which is impossible. As a small dose of reality, previous Linux schedulers had never gone unchanged for long periods, and CFS was already being tweaked before Con wrote BFS.
I cannot take your word for it. You would have to prove this with public posting on mailing list, forum, etc. showing when the work first began (before or after BFS - with posted date).
Just saying this was being worked on before BFS proves nothing. Just your word. And if this had started before Oct 2009 then this 200 line change took a very long time to get done - over one year.
So, either you’re right and Linux developers were slow getting this out or I’m right and it was a response to BFS threat resulting in changing and improving CFS.
Linus also had an issue with Alan Cox:
I only read Linus’ email which accused Alan of introducing a bug into the kernel. So, it goes to show that some Linux developers are pushing bugs into Linux too.
There is no “BFS threat”. It is normal for the Linux scheduler to see tweaks from time to time, and this has happened to CFS over its lifetime as it did to its predecessor. You can easily look for yourself at the history of the files that make up the scheduler and see that it doesn’t go years at a time without improvement as you seem to imagine.
This patch, the “200 line” patch, is new. Someone only recently had the inspiration to try it. It will make no difference whatsoever for a lot of people, despite the excitable Phoronix user who compared it to a completely different scheduler (BFS). It only matters if you use multiple virtual terminals. It relies on cgroups, which are a feature that did not exist when CFS was originally developed. And let me emphasise this again: It has nothing to do with BFS.
You seem to have imagined a lot of stuff that didn’t happen. What I’d suggest is, rather than “reading between the lines” and imagining this big conflict with “threats” you could find out what’s really going on. You might even learn something that could be applied to Haiku (Hint: Haiku could do with a decent scheduler).
Yes, I have to agree with NoHaikuForMe (never thought I’d write that sentence ) This has nothing to do with BFS, and while I’m not going to take sides when it comes to Linus and Con Kolivas, their spat has nothing to do with this. Furthermore Linux is not aimed as a ‘Desktop’ OS, unlike Haiku/Beos which is directly aimed at the desktop. This means that in many (majority?) of situations that Linux is used, throughput is the most important factor. For a desktop OS on the other hand, interactivity is paramount. Now, in order to maximize one you have to sacrifice the other. For an os that wears many hats like Linux, this is a delicate balance. For an OS like Haiku that is directly aimed at the desktop it’s not really much of an issue.
Where is your proof? This is the 2nd time I have asked for you to prove what you say. I gave two possibilities with both having equal chance of being right but you dismissed them. In case 1, this patch was made before BFS showing it takes Linux developers lots of time to make a small patch (over one year). In case 2, it was made in response to BFS to improve latency. Meaning BFS had to push a developer to create this patch in the first place and only exists because of BFS. I guess this paints Linux badly and why it hit a nerve with you!
Since you feel so confident then back it up. Where are the mailing list posts? Where are the forum posts? Where are the commit logs? That’s right, you can’t provide them because maybe you’re in the wrong and don’t want anyone to know.
Hmmmmm, you are just good at yapping on and on.
2nd, BFS was meant to reduce latency and improve responsiveness for multi-tasking. This patch does the EXACT SAME THING FOR CFS. I highly doubt this to be a coincidence.
3rd, if CFS was in the works to get this performance patch first then why did Con waste his time developing BFS? Only makes sense if CFS lacked responsiveness, with no immediate patch, when BFS was being developed. We call this using logic and reasoning.
Apparently you are super biased towards Linux and only can see your version of events. Any negativity for Linux and you start flipping out. One of my cases is right and show a big problem with Linux development which some Linux supporters would rather hide or ignore.
"And let me emphasise this again: It has nothing to do with BFS."
Show me a post from the developer of the patch (Mike Galbraith) to prove this was being developed before BFS release!!!
Linux is for desktop + server markets so not sure where you get your info from Rox. Who do you think all those distros are for? Like Ubuntu, Debian, Fedora, etc.
I am happy to see these issues being fixed but makes you wonder what is going on with Linux development to happen so slowly. All those devs and still takes time to get bugs fixed and improve performance.
You are offering a false dichotomy. A false dichotomy occurs when the arguer excludes the real situation by insisting that only one of two possibilities can be true, and all others are arbitrarily excluded. Let me give you another example, so that you can see how silly this is. Suppose I insist that either you, tonestone57 are a Martian, or you are a vampire. Which is it?
[quote]
2nd, BFS was meant to reduce latency and improve responsiveness for multi-tasking. This patch does the EXACT SAME THING FOR CFS. I highly doubt this to be a coincidence.[/quote]
I think that’s the root of your problem. What you’ve described is one of the goals of a scheduler design. Essentially your claim amounts to “BFS was a good scheduler, therefore any other scheduler that claims to be good, or any change to a scheduler that improves it, is inspired by BFS”. But this is obviously ridiculous.
Your mistake here is named ‘post hoc ergo prompter hoc’. Which means roughly “before, therefore because”. Just because this patch, and many others, were developed chronologically after BFS, does not make them related to BFS. I have already provided a non-technical example which illustrates this mistaken way of thinking. “Breakfast at Tiffany’s” precedes “Star Trek”, but it did not cause Star Trek to be created. There were also patches to improve the scheduler before BFS, and before CFS, and Con himself wrote at least three other schedulers with the goal of improving interactive performance. Tweaks to the scheduler will continue forever, it’s not one of those things you just write once and forget about.
Forever. All operating systems have “a responsiveness issue”. The user wants everything now, and that isn’t possible. So beyond a certain point, they will be disappointed. The more users can do (and in Linux they can do a whole lot) the more opportunities there are to disappoint them. So the job of trying to improve this is endlesss (note: not pointless, but without end)
This seems to be central to your misunderstanding. Responsiveness isn’t like addition, where you can learn how to do it, and from there on it’s plain sailing. You don’t get to say “OK good, we don’t have any responsiveness problems in our OS, so now we can move on”. That never happens.
Which reminds me, you might want to check that “R1 feature list” they’re developing for Haiku. It would be a shame if “actually write a decent scheduler” was missing from the list after it was a centre point of the argument in favour of spending many years writing a bespoke kernel for Haiku, wouldn’t it?
[quote=tonestone57]Linux is for desktop + server markets so not sure where you get your info from Rox. Who do you think all those distros are for? Like Ubuntu, Debian, Fedora, etc.
[/quote]
Yes, I wrote Linux is an OS that ‘wears many hats’, as in has many uses, from running in toasters to huge computer clusters. Haiku on the other hand is a desktop OS, it will not perform in the same league as Linux when used as a server for instance, the reason for this is that it sacrifices throughput for responsiveness under heavy load. What is your point exactly?
This actually describes the only responsiveness problems I’ve had with Linux (and it’s my primary OS, I use it every day, though I hope that in the future Haiku will bump it down to secondary), which is disk I/O related. When I do very disk-heavy operations in Linux the gui will start to stutter, this is again the priority of throughput vs responsiveness. For a server and many other situations you want the disk operation to finish as fast as possible, for a desktop OS you will sacrifice disk operation speed in order to have a fully responsive system in the meantime. There’s no magic involved, but there is a balance which can be hard to get right if there’s no obvious target (like with Haiku that directly targets desktops). As for responsiveness during heavy cpu usage I’ve encountered no problems, here Linux performs perfectly for me.
You also wrote: “Furthermore Linux is not aimed as a ‘Desktop’ OS"
and "unlike Haiku/Beos which is directly aimed at the desktop”
Maybe you forgot to quote that?
Linux is aimed at desktop usage as well as server and other devices too.(ie: mobile devices and embedded devices like cell phones, routers, gps, etc.). Android has displaced Linux on cell phones and tablets becoming more popular and used more as the “free” alternative.
Linux has been pushed for the desktop since the very start. Most of the distros are for desktop use. You make it sound like Linux only targets servers mainly and desktop (& embedded) use is just a side effect. Linux has been targeting the desktop but has gained little market share there. Targeting a market and getting acceptance/popularity are two different things.
Example: I can sell a product to the elderly (targeting) but doesn’t mean they will buy and like my product (popularity; % of market share)
My point is Linux is targeted for desktop also. I also would think responsiveness as being the most important to a desktop OS which Linux is starting to finally realize with those changes/patches + Wayland. Linux has worked good for me but never great or excellent. Those improvements should finally make it better than just good. I also found it interesting to see why these improvements came along and how long they took to get incorporated (ie: don’t happen right away - more like a year or longer).
Oh please, ‘pushed’?. The only serious push I’ve seen for Linux on the desktop has been Ubuntu, and that push has gotten good results. Beos in contrast was a commercial OS directly targeting the desktop and written to do so from the ground up which meant integrated GUI, integrated sound system, integrated desktop environment etc.
Yes, the desktop use is a side effect (embedded devices is NOT, it’s a huge market for Linux), Linux is a KERNEL. It doesn’t come with anything that would make it a desktop OS. People have later added things to make it one, so we have tons of window managers, desktop environments, sound architectures, etc etc at different stages of maturity, this is because it was NOT intended as a desktop OS with a GUI and all bells and whistles, UNLIKE Beos which was exactly that. The distros is what YOU see, but that’s not where the big money is being made with Linux, the desktop is NOT why Microsoft sees Linux as a competitor it wants to get rid of. Heck, Con Kolivas quit the Linux kernel development because he was frustrated with the lack of focus on the desktop and the favoring of server performance, which in his opinion was the reason his patches were rejected.
That Linux works as well as a desktop OS kernel as it does is a testament to it’s ability to be used in so many different settings, but it will never have the same integration as an OS directly aimed for the desktop, like Beos or Windows.
Linux main use is NOT that of desktops, it’s in servers, and embedded systems like routers, media players, smart phones, microwave ovens, media players, etc etc, that’s were the money is and that’s were the development money has been spent.
Maybe you need to wake up, A) Again, apart from Ubuntu there has been precious little money put into the ‘Linux Desktop’, B) As for market share, boy if Haiku ever reaches the current desktop market share of Linux I will be ecstatic, but being a realist I seriously doubt it.
Again it’s so very simple, the kernel devs (AGAIN, Linux is a KERNEL) have had precious little interest in improving Linux as a kernel for a DESKTOP OS since the money in Linux lies elsewhere (servers, embedded devices).
I’m guessing you are simply someone with a grudge against Linux for some reason and now you are simply trolling.
Rox wrote:“I’m guessing you are simply someone with a grudge against Linux for some reason and now you are simply trolling.”
I’m the one that said Linux is targeting and being used for desktop OS and I’m trolling? Look again. You’re the one that is trolling saying Linux is not for desktop when even Wikipedia agrees with me and you went on with this big rant to say you were right about this point when you clearly weren’t. Showing you have strong bias and feelings on the topic.
Yes, Linux has taken off on server market, everyone knows that but the distros are for desktop (individual use). Something you fail to see yourself. Who do Ubuntu, Fedora, OpenSuse, Sabayon, PCLinuxOS, Mandriva, etc. target? Desktop users. Yes, these distros can be used for servers too but they were created for the average computer user to run at home.
“Linux (commonly pronounced /ˈlɪnəks/ LIN-əks in American English,[4][5] also pronounced /ˈlɪnʊks/ LIN-ooks[6] in Europe and Canada) refers to the family of Unix-like computer operating systems using the Linux kernel.”
“Typically Linux is packaged in a format known as a Linux distribution for desktop and server use.”
The information I provided about Linux is accurate and true! I have not made it up like you have about not marketing as desktop OS. I even gave links to support my comments where possible and even 2 cases for making of the CFS patch; one of which is right, but maybe reading the links I gave is too hard for you? Linux had responsiveness issues - 2 different situations. Patches were made to fix these issues. I pointed the issues/patches out with links. Maybe you wanted to sweep these issues under the rug and hide them? X has performance issues too. Why else are developers working on Wayland to replace it? I can pretty much prove everything I said so how is that trolling? Just because you disagree with me? I gave true information and supportive links. Maybe you’re not used to technical analysis? These issues were real and recent. I just didn’t make them up for the fun of it.
I am very happy that Linux is now looking at improving and fixing OS responsiveness and focusing equally on desktop use. Sounds like you’re the one upset about this. Haiku will have a tougher time competing with Linux because of changes like these which improve latency and performance.
And what is a linux distro? 99% of them aren’t even adding anything to the Linux software stack, they just take what is there and combines it in different configurations and call it distro X. And since noone is making money on the desktop there hasn’t been any major ‘push’ as you describe it until Ubuntu, which I seriously doubt is anywhere near breaking even with what Mark Shuttleworth has poured into the project. Again as it stands the money in Linux lies in servers and embedded devices, which is where the effort has been placed. And since you like to qoute wikipedia:
‘Primarily based on web server statistics, various companies estimated that the desktop market share of Linux range from less than 1% to 4.8%.[80] In comparison, Microsoft operating systems hold more than 85%.’
‘In September 2008 Microsoft CEO Steve Ballmer admitted that 60% of web-servers run Linux versus 40% that run Windows Server.’
Does this paint a picture? Linux is PUSHED for the server and embedded market, and even with Ubuntu the resources for improving the linux desktop experience are miniscule by comparison.
So stop staring yourself blind on the number of distros which are 99% nothing more than a combination of existing packages with a volunteer community behind it. We could have that for Haiku aswell if it starts gathering attention but having 20 distro’s won’t make Haiku in itself any better, only code contributions will.
What I love with Haiku/Beos is that it’s modeled to be a desktop/GUI os from top to bottom, everything nicely integrated with standard api’s for gui, sound etc. And this is something I think Linux will never achieve even if it get’s a major desktop push, since the pieces that make up the ‘Linux desktop’ are from very different origins and often doesn’t fit very well.
Beos/Haiku is the best desktop environment I’ve ever used and since I’m primarily a desktop user it’s what I look forward to using it as my primary os once it has matured sufficiently. Until then I will get my work done on Linux.
“And what is a linux distro?”
Distros are versions of Linux OS for desktop and server use.
A couple of the distros began back in 1994-1996. Like, Debian, Slackware, Redhat and aimed to capture the desktop market from back then.
“Debian can be used as a desktop as well as server operating system.”
“And since you like to qoute wikipedia:
‘Primarily based on web server statistics, various companies estimated that the desktop market share of Linux range from less than 1% to 4.8%.[80] In comparison, Microsoft operating systems hold more than 85%.’”
The key phrase being:
“various companies estimated that the desktop market share of Linux”
In case you missed it, the keyword was: desktop, meaning it has a presence in the desktop (home) market.
Very important! Since you went to Wikipedia, did you not notice on the right hand side the summary box where it says: “Marketing target - Desktops, servers, embedded devices”
“Does this paint a picture? Linux is PUSHED for the server and embedded market…”
Linux is just popular for server and embedded use. I explained the difference between marketing towards (or targeting) a specific market versus popularity in that market. They are not the same! What percent of the desktop market would Linux have to capture to be considered a desktop OS by you? 5%? 10%? 20%? 30%? 50%? You can’t use market share to say it was targeted for that market because market share only gauges popularity in different markets.
Example: what if I create a muffler for trucks and cars and I get 40% truck market and 2% car market. Does that mean my muffler is only for trucks? According to you, it does. But in reality the muffler is made for and targeted to both markets.
I agree that Linux is made up of many different components that don’t mesh as well as Haiku. You have one set of developers doing the kernel, others on X, others on Gnome, KDE, fluxbox, etc. But, that doesn’t exclude the OS from targeting desktop use. Only means Linux is not tightly integrated like Haiku. You are looking at how Linux is put together and not what markets it is targeted towards.
What you don’t understand is:
Linux is targeted for desktop, server and embedded use (Wikipedia agrees - look @ summary box)
Linux is popular for server & embedded use and unpopular for desktop use These are two different things which you keep confusing! Both true, but I am talking about #1 and you are trying to use #2 to say #1 is not true. You also throw in a #3 saying Linux is a bunch of parts put together and so does not qualify as a desktop OS. Does not matter how Linux is put together because it still markets to desktop users.
“I will get my work done on Linux.”
You say Linux is a server OS but you use it @ home on desktop. I’m surprised you don’t use a desktop OS like Windows or MacOS X then if Linux is only for server and embedded like you say.
If I run linux it’ll be on a server. Thats it. Its otherwise just as practically useless as BEOS,MAC OSX,HAIKU etc.
Theres a reason MS dominates the desktop. they have all the applications and all the drivers.
None of the freeware on the market is competitive. If you add up the market share of linux,macosx,haiku,bsd etc etc etc, the make up like 8% of the market for dekstop OS.
As for embeded stuff with linux. I expect ARM to get handed its ass on tablet and then phones. INTEL and AMD want some market share.
) This has nothing to do with BFS, and while I’m not going to take sides when it comes to Linus and Con Kolivas, their spat has nothing to do with this. Furthermore Linux is not aimed as a ‘Desktop’ OS, unlike Haiku/Beos which is directly aimed at the desktop. This means that in many (majority?) of situations that Linux is used, throughput is the most important factor. For a desktop OS on the other hand, interactivity is paramount. Now, in order to maximize one you have to sacrifice the other. For an os that wears many hats like Linux, this is a delicate balance. For an OS like Haiku that is directly aimed at the desktop it’s not really much of an issue.