Announcement

Collapse
No announcement yet.

Fedora, Debian, FreeBSD, OpenBSD, OpenSolaris Benchmarks

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #41
    Originally posted by kraftman View Post
    You didn't know what were you talking about.
    Maybe you didnt explain anything? You wrote "different ram amount". What did you mean with that? Now I understand, but what is clear to you, is not clear to others.

    Originally posted by kraftman View Post
    And this means Linux CPU utilization must be higher in this benchmark? You're really making fool of yourself.
    I meant that Linux used faster CPUs and faster RAM. And still Linux had lower SAP score than Solaris. And we also see that Linux had 87% utilization, whereas Solaris had 99%. What conclusions can you infer from these facts?

    Originally posted by kraftman View Post
    At your first arrival you said Linux doesn't scale good on Big Irons. I showed you many times it does
    You did??? Can you post them links again? I must have missed them!

    Originally posted by kraftman View Post
    So what are you trying to prove here? Sure, they used 128 blades with 256 CPUs and, for instance, achieved 9,6 million BOP/s in SPECjbb2005. Which is quite respectable.

    But the SPECjbb2005 benchmark (configured with multiple JVMs) is one of the problems of the "embarrassingly parallel" class. You can scale it by running multiple copies. In the bench, they configured 128 JVMs, exactly the number of blades. The application didnt have to use NUMA link between the blades. The JVMs are working independent from each other, so no communication between the the nodes, too. So the SPECjbb2005 benchmark on an Altix 4700 is not much more than throwing a cluster at a problem it's best at! Hence, this does not prove anything about scalability.

    We all agree (including Linux scalability experts), Linux scales well on a cluster. That is where Linux strength is, horizontal scaling. As Linux scalability expert Greenblatt says:
    "Greenblatt: Linux has not lagged behind in scalability, [but] some [Unix] vendors do not want the world to think about Linux as scalable. The fact that Google runs 10,000 Intel processors as a single image is a testament to [Linux's] horizontal scaling".

    "The true Linux value is horizontal scaling."

    To my knowledge, there is no single big machine with 10.000 intel processors on the market. Google must be using clusters.

    Originally posted by kraftman View Post
    Show me Solaris scaling on at least 512CPU machine not some stupid sap papers.
    Why are the SAP papers stupid? Is it because Linux looses? If Linux win, then SAP papers are good, yes? So what is a good paper? A paper is good if Linux wins, and bad if Linux looses?

    To me, a paper is good if it uses official standardized benchmarks, in a fair way, where no one complains. The outcome does not decide if a white paper is good or not. If the white paper has a scientific, reproducible approach, then it is good. Just like in science. The method is extremely important, the outcome is not important in science.

    But you have proved earlier that the outcome decides if a white paper/benchmark is good, not how good the method is. I think you(?) showed earlier that one guy migrated from 800MHz SPARC to 2.4GHz Intel dual core Linux and therefore you "proved" that Linux is faster. That is just a weird comparison according to me. Not fair. But that doesnt matter to you, as you have "proved" what you want.

    I want more benches on the SAME hardware. With many CPUs/cores. But I suspect you dont want such benchmarks, because Linux would loose big time. You prefer unfair benches. That is not fair.



    Regarding Solaris scaling to 512 CPUs, I dont know if there exist such a Sun machine on the market, so I can not show you such benches. But if there existed such a machine, Solaris would of course have higher CPU utiilization, again. Why wouldnt Solaris?

    You know, just because Linux EXISTS on a big machine, it doesnt mean that Linux scales well. I bet that Linux has a very low utilization on machines with many CPUs, maybe 10% CPU utilization? But I am sure that Solaris utilizes the CPUs far better than Linux on same hardware (machines with many CPUs). As we saw on the SAP benchmark with 48 cores.

    Originally posted by kraftman View Post
    Btw where [Larry Ellison] said Solaris is better?


    Originally posted by kraftman View Post
    [Larry Ellison] means "we can make some money on it if we'll continue Sun's propaganda, because we've got slowlaris while our competitors, like IBM, don't".
    Heh. Dont you know that you need four IBM POWER6 5GHz to match two Intel Nehalem 2.93GHz on TPC-C benchmarks? Dont you know that any modern AMD/Intel CPU is 5-10x faster than a modern IBM Mainframe CPU? Dont you know you need six IBM Power P570 servers with 28 POWER6 CPUs to match one Sun T5440 (that has four 1.4GHz Niagara CPUs) on SIEBEL v8 benches? One P570 costs $413.000 and one Sun T5440 costs $76.000. Dont you know that Linux has the TPC-C record right now, on SUN hardware? IBM has lost it. To me it seems that IBM is slower everywhere.

    Comment


    • #42
      Originally posted by kebabbert View Post
      Maybe you didnt explain anything? You wrote "different ram amount". What did you mean with that? Now I understand, but what is clear to you, is not clear to others.
      You're asking here I should treat like an idiot? It was clear also for some another person, it seems it's only you who has problems with understanding things.

      I meant that Linux used faster CPUs and faster RAM. And still Linux had lower SAP score than Solaris. And we also see that Linux had 87% utilization, whereas Solaris had 99%. What conclusions can you infer from these facts?
      And Solaris used different DB and twice as much memory as Linux. Conclusion is very simple - on Linux CPU utilization was 87% using different DB and less memory and on Solaris CPU utilization was 99% using different DB and double amount of memory. That's all.

      You did??? Can you post them links again? I must have missed them!
      I asked for another thread and you ignored this like a troll.

      So what are you trying to prove here? Sure, they used 128 blades with 256 CPUs and, for instance, achieved 9,6 million BOP/s in SPECjbb2005. Which is quite respectable.
      Not trying, but I actually proved you Linux scales great.

      But the SPECjbb2005 benchmark (configured with multiple JVMs) is one of the problems of the "embarrassingly parallel" class. You can scale it by running multiple copies. In the bench, they configured 128 JVMs, exactly the number of blades. The application didnt have to use NUMA link between the blades. The JVMs are working independent from each other, so no communication between the the nodes, too. So the SPECjbb2005 benchmark on an Altix 4700 is not much more than throwing a cluster at a problem it's best at! Hence, this does not prove anything about scalability.
      Straw man strikes again. You can safely ignore mentioned benchmark if you want, but it doesn't change the fact that Linux scales great on machines were Solaris probably didn't even try. You should at least point to source of your quoted text, but I already found it:

      At the moment i have my problems to see why both benchmarks highlight the advantages that make them a better choice for Java. Perhaps BlueToTheBone should stop to use those numbers as examples to convince the world of this opinion in regard of Sun.
      http://www.c0t0d0s0.org/archives/593...benchmark.html

      Someone from OpenSolaris user group. I said many times I'm not interested in SUN related people bull. Btw. you pasted someones text here like it was yours.

      We all agree (including Linux scalability experts), Linux scales well on a cluster. That is where Linux strength is, horizontal scaling. As Linux scalability expert Greenblatt says:
      "Greenblatt: Linux has not lagged behind in scalability, [but] some [Unix] vendors do not want the world to think about Linux as scalable. The fact that Google runs 10,000 Intel processors as a single image is a testament to [Linux's] horizontal scaling".

      "The true Linux value is horizontal scaling."

      To my knowledge, there is no single big machine with 10.000 intel processors on the market. Google must be using clusters.
      What knowledge? It's about horizontal scaling and what you were talking all the time was about vertical scaling.

      Why are the SAP papers stupid? Is it because Linux looses? If Linux win, then SAP papers are good, yes? So what is a good paper? A paper is good if Linux wins, and bad if Linux looses?
      Read links I showed you.

      I want more benches on the SAME hardware. With many CPUs/cores. But I suspect you dont want such benchmarks, because Linux would loose big time. You prefer unfair benches. That is not fair.
      Funny to read this. It is you who's giving unfair benches - different memory amount, different db, SUN made benchmarks. Thanks for such bull.

      Regarding Solaris scaling to 512 CPUs, I dont know if there exist such a Sun machine on the market, so I can not show you such benches. But if there existed such a machine, Solaris would of course have higher CPU utiilization, again. Why wouldnt Solaris?
      If you can't then what are you trying to do?

      You know, just because Linux EXISTS on a big machine, it doesnt mean that Linux scales well. I bet that Linux has a very low utilization on machines with many CPUs, maybe 10% CPU utilization? But I am sure that Solaris utilizes the CPUs far better than Linux on same hardware (machines with many CPUs). As we saw on the SAP benchmark with 48 cores.
      Like I said, meaningless comparison. It wasn't same hardware, so stop lying.

      I read this, but I don't see where he said such thing. Of course he's doing what he said before:

      Ellison insisted that Oracle can continue to give full support to the Linux operating system, even as it gears up to promote Solaris, the operating system developed by Sun.

      Comment


      • #43
        Originally posted by kraftman View Post
        You're asking here I should treat like an idiot? It was clear also for some another person, it seems it's only you who has problems with understanding things.
        I am just asking you to be clearer by providing more information and relevant links. Is that hard for you to do? I have asked for this over and over again. But still you dont. You just write very terse answers making it hard for me to follow you. The thing is, I think very differently from you, so I am getting confused and I dont understand what you mean. Because you state so weird things sometimes, so I have a hard time understanding what you say "did Kraftman really compare 800MHz SPARC to a 2.4GHz Intel? No, he could not possibly mean it is a fair comparison! Or? What did he mean? I dont get it".

        I just dont understand you. Therefore I have asked you to be much clearer. But apparently that is very very very difficult for you. Because otherwise you would not make such a big deal of insisting of being terse.

        For instance, you have posted several links about Oracle and SAP, like this one

        Actually, I dont understand why posted that link. You just posted several such links without an explanation. What did you mean with them? No information at all, from you on this.

        Look, is it really really really sooooo difficult to provide some more information? I have asked this many times. I dont understand what you are trying to say!

        So, if this is the only thing that helps: Yes, please treat me like an idiot. Please explain what all those links where about. And when you explain something, please provide some explanation. Can you do that, to an idiot? (Obviously you refuse to provide information when I ask of you)


        Originally posted by kraftman View Post
        And Solaris used different DB and twice as much memory as Linux. Conclusion is very simple - on Linux CPU utilization was 87% using different DB and less memory and on Solaris CPU utilization was 99% using different DB and double amount of memory. That's all.
        So no conclusion can be drawn at all? You dont think that any conclusions can be drawn at all?


        Originally posted by kraftman View Post
        I asked for another thread and you ignored this like a troll.
        Why should I do what you order me? You attacked me when I said that they should try a new release of OpenSolaris, that is all I said. And you called me Troll for that post. So, if you really want another thread, why dont you start it, instead of ordering me to do it, and calling me a Troll? Do you often do like that? Command people to do something that YOU want, and when they dont do it, you call them names?


        Originally posted by kraftman View Post
        Not trying, but I actually proved you Linux scales great.
        You did? Cool! Can you show me those posts, then? You showed that Linux got high scores on parallell work loads with 256 CPUs. But I must have missed your other links. Please treat me like an idiot and repost them here. Can you do that, or is it too difficult (as usual)?


        Originally posted by kraftman View Post
        Straw man strikes again. You can safely ignore mentioned benchmark if you want, but it doesn't change the fact that Linux scales great on machines were Solaris probably didn't even try. You should at least point to source of your quoted text, but I already found it:



        Someone from OpenSolaris user group. I said many times I'm not interested in SUN related people bull. Btw. you pasted someones text here like it was yours.
        Yes, that is the link I got this text from. Earlier I always posted links, but I stopped because of you. Recently, I showed you a Sun page which pointed to these certified SAP white papers on www.sap.com but you explained that these SAP certifications on www.sap.com had been pointed to, from a Sun web page, and therefore these SAP certifications are "Sun propaganda" and not to be trusted. I said that from now on, I will not link to Sun pages, I will copy & paste text. Do you remember this discussion? Therefore I dont provide Sun links anymore.

        Back to this SPECjbb2005 benchmark. Do you mean the things I wrote down are not correct? So please, point out the incorrect facts here, then. Exactly what is wrong? Or dont you know what is wrong, or maybe everything is correct?

        "But the SPECjbb2005 benchmark (configured with multiple JVMs) is one of the problems of the "embarrassingly parallel" class. You can scale it by running multiple copies. In the bench, they configured 128 JVMs, exactly the number of blades. The application didnt have to use NUMA link between the blades. The JVMs are working independent from each other, so no communication between the the nodes, too. So the SPECjbb2005 benchmark on an Altix 4700 is not much more than throwing a cluster at a problem it's best at! Hence, this does not prove anything about scalability."


        Originally posted by kraftman View Post
        What knowledge? It's about horizontal scaling and what you were talking all the time was about vertical scaling
        I dont understand this. Could you explain what you mean? I said that Linux doesnt scale well vertically, but scales well horizontally. Just as the Linux scaling experts say.


        Originally posted by kraftman View Post
        Read links I showed you.
        How about you explain a bit, instead? I asked if a benchmark is good if Linux wins, but bad if Linux looses? And to this question you answer, very cryptical: "read links I showed you". Which links, you have posted many links. Can you just explain a bit more? It does not help to ask you of more information, maybe I can get more infomation if you treat me like an idiot. So go ahead, treat me like an idiot. I dont get it, why, oh why, is it so difficult to explain exactly what you mean, so we can follow you? I have asked you many times of this.


        Originally posted by kraftman View Post
        Funny to read this. It is you who's giving unfair benches - different memory amount, different db, SUN made benchmarks. Thanks for such bull.
        So you mean that because Sun made benchmarks therefore you can not trust them? So... who do you think should do the Solaris benchmarks? Should FedEx do the Solari benches?

        You know, Linux used faster RAM and faster CPUs. The hardware is in favour of Linux, but still Linux is slower on SAP. So we can conclude that Linux is not a good SAP performer, right? You should use Solaris for SAP. Do you agree on this? Linux sucks on SAP - agreed?


        Originally posted by kraftman View Post
        If you can't then what are you trying to do?
        The whole discussion is which OS scales best. If Linux EXISTS on a machine with 256 CPUs, doesnt say how good Linux scales, right? I hope everyone understands this. But, I have showed that Solaris has higher CPU utilization than Linux on 48 cores, on SAP. Solaris scales better on 48 cores. The more cores, Solaris will be faster and faster than Linux.


        Originally posted by kraftman View Post
        Like I said, meaningless comparison. It wasn't same hardware, so stop lying.
        Ok, so if I install Solaris on a machine with 10,000,000 CPUs, and then I start one SPECjbb2005 JVM on each CPU, and run them in parallell, and then I get lots of BOP - does this mean that Solaris scales better than buggy Linux? No? But you are using the same argument here. It doesnt matter if Linux exist on a computer with the slow Itanium CPU, with 128 blades - it does matter how good Linux utilizes the CPUs, how good it scales. These are two different things. Linux existing on a large computer, does not mean it scales well. As we have seen, Linux underperfoms badly on as little as 48 cores. I expect Linux to be really bad on many cores. CPU utilization will drop off sharply. 87% on 48 core, maybe 75% on 64 cores, maybe 50% on 128 cores. On 512 cores, maybe 10%.


        Originally posted by kraftman View Post
        I read this, but I don't see where he said such thing. Of course he's doing what he said before:
        Larry Ellison says Solaris is better than Linux in that link. I copy and paste for you:
        "We're big supporters of Linux, [but] Solaris is an older and more capable operating system," Ellison said.

        While he expects to see Solaris primarily at the high end, it will go all the way down to the desktop for development. Nevertheless, he stressed that the high end is the home for Solaris, which could be a cloud of x86 or SUN SPARC machines.

        "We think it will be a long time before Linux ever catches up," Ellison said"

        Linux is for low end, and Solaris for high end. Just as it should be.



        And you never answered to my questions: what does Linus T mean when he says that the kernel is bloated? What does Andrew Morton say when he says that the code quality is declining? What does the kernel developers mean with that?

        Comment


        • #44
          Originally posted by kebabbert View Post
          "We think it will be a long time before Linux ever catches up," Ellison said"

          Linux is for low end, and Solaris for high end. Just as it should be.
          I won't feed you more in this thread. Linux handles many mission critical systems and most demanding workloads, so you're just making fool of yourself. You never started a new thread when asked, so your intention is to pollute this thread. It's enough for me to know Orvar Korvar is a big troll.

          Look, is it really really really sooooo difficult to provide some more information? I have asked this many times. I dont understand what you are trying to say!
          I provided you many examples, but it's not my problem you don't understand a thing.

          As some example look here:

          The whole discussion is which OS scales best. If Linux EXISTS on a machine with 256 CPUs, doesnt say how good Linux scales, right? I hope everyone understands this.
          It exists even on bigger machines - 512, 1024CPUs. It scales up to 4096CPUs and Solaris probably don't even scale up to 512CPUs. It says Linux scales up to far more CPUs then Solaris, so Linux scales better.

          But, I have showed that Solaris has higher CPU utilization than Linux on 48 cores, on SAP. Solaris scales better on 48 cores. The more cores, Solaris will be faster and faster than Linux.
          You showed two different papers where Linux and Solaris ran on different hardware. That's all.

          And you never answered to my questions: what does Linus T mean when he says that the kernel is bloated? What does Andrew Morton say when he says that the code quality is declining? What does the kernel developers mean with that?
          I did, long ago.

          Comment


          • #45
            Originally posted by kebabbert View Post
            what does Linus T mean when he says that the kernel is bloated?
            In the very same interview he said kernel is bloated and huge but there is not problem with stability.
            "I think we've been pretty stable,"
            "We are finding the bugs as fast as we're adding them ? even though we're adding more code."

            So you imply that Linux is buggy because it is bloated, while Linus in the very same post you mentioned says this is not a fact.
            So while Linux is bloated and huge, is not buggy. So the problem of bloatness has to do with difficulty to maintain a large amount of code (fortunately kernel has many developers) andmainly a large amount of unecessary code, which while it's still bad doesn't affect Linux' performance.

            Comment


            • #46
              Originally posted by Apopas View Post
              In the very same interview he said kernel is bloated and huge but there is not problem with stability.
              "I think we've been pretty stable,"
              "We are finding the bugs as fast as we're adding them ? even though we're adding more code."

              So you imply that Linux is buggy because it is bloated, while Linus in the very same post you mentioned says this is not a fact.
              So while Linux is bloated and huge, is not buggy. So the problem of bloatness has to do with difficulty to maintain a large amount of code (fortunately kernel has many developers) andmainly a large amount of unecessary code, which while it's still bad doesn't affect Linux' performance.
              And Linus was according to some old, previous Linux' versions.

              Comment


              • #47
                Bloated compared to Linux 2.x code lines etc. Some people don't understand this:

                No one would argue with Linus on the status and well being of the kernel, but it’s so easy to take “bloated” out of context. How does it compare to the Windows or BSD kernels? He’s not saying that the kernel is bad or inefficient, just that it’s not as neat and clean as it perhaps could, or should be.

                Comment


                • #48
                  Originally posted by kraftman View Post
                  I won't feed you more in this thread. Linux handles many mission critical systems and most demanding workloads, so you're just making fool of yourself.
                  We all agree that Linux handles many critical systems and demanding workloads. No one questions this claim, it is true. Yes, Linux is good enough to do that. But Linux is not best. There are better OS out there. Or do you rank Linux above IBM z/OS Mainframes, HP-UX, Sun Solaris, IBM AIX, OpenVMS? Do you think Linux is the most stable OS there is? There is no better? You do not agree that there are better OSes than Linux?

                  I am not saying that Linux is bad, it is not. It is for instance, far better and more stable than the best Windows version. But Linux is not that perfect that every Linux zealot seems to believe, it has it's flaws. Linux is not flawless. There are better OSes out there. Not counting Solaris, I would use Linux only. I use Ubuntu at home, and like it. But, Linux has some bugs and is not that stable, alas.

                  Personally, I think Linux is very good for Unix. Because if Linux did not exist, everyone would use Windows. From Windows, the step to Unix is too big. But now, even Windows people runs Linux. And from Linux to Unix, the step is quite small. So, Unix potential user base has grown tremendously, because there are so many Linux users out there.

                  See for instance this company, full of Linux zealots, that switch to OpenSolaris:

                  If they had run Windows, the switch would have been more unlikely.


                  Originally posted by kraftman View Post
                  You never started a new thread when asked, so your intention is to pollute this thread. It's enough for me to know Orvar Korvar is a big troll.
                  I wrote that Phoronix should try the latest OpenSolaris build in this benchmark. For that, you attacked me, called me a Troll. And I had to defend myself, or do you rather I be quiet so you can call others name, freely? And then you continued harassing me. And I defended. And you continued. Then, suddenly you say that I should start a new thread. Maybe you wanted to hide all your harassment in another thread? You thought not many people will read another thread, but many reads this benchmark? And I said YOU can start YOUR thread, if you wish. Because I refused to obey you, you say that I "want to pollute this thread" and call me "big troll". Not really fair, to me.

                  Everything is written here, recorded in the posts. Just read them in this thread, and see if I speak true, or false.


                  Originally posted by kraftman View Post
                  I provided you many examples, but it's not my problem you don't understand a thing.
                  I've told you that we think differently. I can not follow your thoughts, you need to explain them to me. I have said this to you, many a times. To no avail. You still refuse to explain what you mean. Why do you refuse to explain what you mean, instead of "read all links, and you will understand what I mean"? What links, there are no links! You have posted many links earlier, but which of them do you refer to? And what do you want me to understand? You dont explain what you are trying to tell me. I have no clue.


                  Originally posted by kraftman View Post
                  It exists even on bigger machines - 512, 1024CPUs. It scales up to 4096CPUs and Solaris probably don't even scale up to 512CPUs. It says Linux scales up to far more CPUs then Solaris, so Linux scales better.
                  First of all, you know nothing of how many CPUs Solaris can scale to. So I find it hard to believe you when you say that "Solaris probably doesnt scale up to 512CPUs" - becuase there is no evidence at all supporting you. In fact, everything speaks against you - Solaris has reputation of scaling well. Linux has not. You have not provided links, nor benchmarks, nor white papers, no evidence at all, to support you. Nothing. If you provided evidence of Solaris scaling bad, then maybe there would be some substance in your claims. But no.

                  Second, if Linux EXISTS on a big computer, it does not mean that Linux scales well. So if I assemble a machine with many many dog slow CPUs, the machine is not even efficient, it is slow and bad - and if I compile Solaris to that bad machine with many CPUs - does that mean Solaris scales better than Linux? The mere existence does not say anything. How many percent CPU utiilization does Linux have, on many CPUs, and how much percent has Solaris? THAT is the interesting question. Not if someone compiled Linux to a big computer - that is totally irrelevant.

                  15 years ago, I read an article by a computer scientist, a famous professor. He said that scalability is very difficult. On 1000 CPUs, the best researchers in the world, could only utilize 12 CPUs fully. You could replace that 1000 cpu machine, with 12 cpus - and get the same work done. Vertical scalability is very very hard, it takes decades. If you expect Linux (which scaled 2-4CPUs in v2.4) to scale to 4096 CPUs in v2.6 - and do that well, then you should read more about parallell computing research. Vertical scaling is very hard to do well. To scale, is easy - it is just a matter of compiling Linux to that big machine and run some parallell benches. But to good scaling, is very hard.

                  Everyone agrees that Linux scales well horizontally. Google runs probably 1000s of Linux nodes in a large cluster. The Linux scaling experts say that "the true value of Linux lies in horizontal scaling". And we all agree with the Linux scaling experts, dont we?


                  Originally posted by kraftman View Post
                  You showed two different papers where Linux and Solaris ran on different hardware. That's all.
                  Not many benches use exactly the same hardware, same software, same compiler settings, etc. Even if the same hardware/software where used, you would argue about compiler settings. Or the branding of the RAM sticks, etc. The fact is, Solaris used slower cpu and slower ram, and still topped out. Do you mean that Solaris won, because of the database was so much better than the DB used in Linux benchmark? Is that the reason that Linux only got 87% cpu utilization (which is not that good from a scaling point of view)?

                  When Phoronix does these benches, where Linux compiles to 64bit binaries with the newest gcc compiler, and OpenSolaris compiles to 32bit binaries with an outdated gcc compiler - you dont complain. Then you think these benches are fair, because Linux wins big time. When you compare Sparc 800MHz to 2.4GHz Intel dual core Linux, then that bench is fair - you think.


                  Originally posted by kraftman View Post
                  I did, long ago.
                  Oh, I must have missed your answer. Can you do a quick recap what Andrew Morton, Linus T, Dave Jones, Alan Cox, etc mean when they say that Linux is buggy, bloated, going to pieces, and code quality is declining? Or is your answer "read the links" - just as usual? Leaving me to figure out, exactly which links, and what you are trying to say?


                  Another question. I dont like FUD or lies. You claim Sun FUDs a lot about Linux. Can you do a quick recap on some of the Sun's FUD here? If Sun really FUDs, then I must criticize that. Normally, Sun plays fair. But here is an example of Sun not playing fair, and I criticize. Sun is comparing a new system to a bad slow HP Itanium system (this is something IBM does all the time, and it is not fair to compare a new system to an old system). Read the comments to see my criticism



                  Originally posted by Apopas View Post
                  In the very same interview he said kernel is bloated and huge but there is not problem with stability.
                  "I think we've been pretty stable,"
                  "We are finding the bugs as fast as we're adding them — even though we're adding more code."
                  To me, "pretty stable", is not stable. And as we have seen from numerous posts and articles, there are lots of problems with stability. I have posted several such links, I can repost them if you wish.

                  Linus claims he finds bugs as fast as they are adding them, even though they add more code. But, you have to remember that Linux code is rewritten all the time. "There is no design, and never will be" - as Linus T said. Linus T says that Linux evolves like biology, try different solutions and keep the best solution. And repeat, slowly Linux is getting better, he claims.

                  Say that they squash many bugs, and then Linus rewrites all that code. So now you have new code, with new bugs. The old bug free code is now gone. So how can you decrease the number of bugs? It is impossible if you have new code every half a year. It takes Service Pack 1 before Windows gets stable enough to use. The first code, is always unstable. Old mature code is stable. This is true.


                  Originally posted by Apopas View Post
                  So you imply that Linux is buggy because it is bloated, while Linus in the very same post you mentioned says this is not a fact.
                  So while Linux is bloated and huge, is not buggy. So the problem of bloatness has to do with difficulty to maintain a large amount of code (fortunately kernel has many developers) andmainly a large amount of unecessary code, which while it's still bad doesn't affect Linux' performance.
                  No, I do not say that Linux is buggy because it is bloated. Your premise is not correct. Bloat is only ONE of the problems that Linux has. Other problems is Linux has no design and therefore is has no stable ABI (your device driver works good, and after an kernel upgrade it breaks or crashes only under special circumstances - how can Linux be stable under such circumstances?). Another problem is that Linux code gets rewritten all the time, to get stable code, it needs to be old and mature. If you squash bugs, and replace that stable code with new, there are new bugs. So there is always new code, and new bugs. Look at Linux sound API. There are 5(five) different sound APIs, last I heard of. Linux sound is horribly broken, people say.

                  Just read what the Linux kernel developers say about Linux kernel. They seem to claim there are many problems.

                  Comment


                  • #49
                    I like the way you replied this time, so I will answer you.

                    Originally posted by kebabbert View Post
                    We all agree that Linux handles many critical systems and demanding workloads. No one questions this claim, it is true. Yes, Linux is good enough to do that. But Linux is not best. There are better OS out there. Or do you rank Linux above IBM z/OS Mainframes, HP-UX, Sun Solaris, IBM AIX, OpenVMS?
                    I rank Linux over Solaris.

                    Do you think Linux is the most stable OS there is? There is no better? You do not agree that there are better OSes than Linux?
                    It's hard to say which OS is the most stable. It probably cannot be proven. I don't consider there are no better systems. For some specialized tasks like real time I'm sure there are better systems now If I remember I didn't say there are no better systems.

                    I am not saying that Linux is bad, it is not. It is for instance, far better and more stable than the best Windows version. But Linux is not that perfect that every Linux zealot seems to believe, it has it's flaws. Linux is not flawless. There are better OSes out there. Not counting Solaris, I would use Linux only. I use Ubuntu at home, and like it. But, Linux has some bugs and is not that stable, alas.
                    Probably nobody denied Linux has bugs, but you were comparing it to Solaris/OpenSolaris and it sounded like SUN systems are more stable, has less bugs etc. This cannot be proven.

                    Personally, I think Linux is very good for Unix. Because if Linux did not exist, everyone would use Windows. From Windows, the step to Unix is too big. But now, even Windows people runs Linux. And from Linux to Unix, the step is quite small. So, Unix potential user base has grown tremendously, because there are so many Linux users out there.
                    That's true.

                    I wrote that Phoronix should try the latest OpenSolaris build in this benchmark. For that, you attacked me, called me a Troll. And I had to defend myself, or do you rather I be quiet so you can call others name, freely? And then you continued harassing me. And I defended. And you continued. Then, suddenly you say that I should start a new thread. Maybe you wanted to hide all your harassment in another thread? You thought not many people will read another thread, but many reads this benchmark? And I said YOU can start YOUR thread, if you wish. Because I refused to obey you, you say that I "want to pollute this thread" and call me "big troll". Not really fair, to me.
                    I didn't attack you because of this. I consider you were trolling many times here and that's why I said what I said. Of course I don't want to hide a thing. Some of your trolling started here:

                    Discuss anything Linux related that doesn't fit into any of the other areas.


                    I've told you that we think differently. I can not follow your thoughts, you need to explain them to me. I have said this to you, many a times. To no avail. You still refuse to explain what you mean. Why do you refuse to explain what you mean, instead of "read all links, and you will understand what I mean"? What links, there are no links! You have posted many links earlier, but which of them do you refer to? And what do you want me to understand? You dont explain what you are trying to tell me. I have no clue.
                    I don't know what you exactly don't understand.

                    First of all, you know nothing of how many CPUs Solaris can scale to. So I find it hard to believe you when you say that "Solaris probably doesnt scale up to 512CPUs" - becuase there is no evidence at all supporting you. In fact, everything speaks against you - Solaris has reputation of scaling well. Linux has not. You have not provided links, nor benchmarks, nor white papers, no evidence at all, to support you. Nothing. If you provided evidence of Solaris scaling bad, then maybe there would be some substance in your claims. But no.
                    It seems it's not documented how many CPUs Solaris can handle running on a single image. However, there are humongous machines which run Linux not Solaris, so this suggest Linux scales better on such machines. I provided you links which shows Linux scales up to many CPUs on a single image. You showed me links which only suggest Solaris scales better in some environments, but those aren't evidences (I agree mine also aren't evidences). You said Solaris has reputation of scaling well and Linux has not. It is known SUN was driving anti Linux campaign since years like mentioned here:

                    http://vger.kernel.org/~davem/cgi-bin/blog.cgi/2007/04/10#bonwick_scalability

                    Second, if Linux EXISTS on a big computer, it does not mean that Linux scales well. So if I assemble a machine with many many dog slow CPUs, the machine is not even efficient, it is slow and bad - and if I compile Solaris to that bad machine with many CPUs - does that mean Solaris scales better than Linux? The mere existence does not say anything. How many percent CPU utiilization does Linux have, on many CPUs, and how much percent has Solaris? THAT is the interesting question. Not if someone compiled Linux to a big computer - that is totally irrelevant.
                    Those SGI machines were very efficient. There are some factors which can affect CPU utilization. In papers you provided there were different Data Base used and different hardware, so I consider you can't judge scalability just on those papers. I've got some different question - does Solaris can handle as many CPUs as Linux? I don't know, but nobody proved if it can.

                    15 years ago, I read an article by a computer scientist, a famous professor. He said that scalability is very difficult. On 1000 CPUs, the best researchers in the world, could only utilize 12 CPUs fully. You could replace that 1000 cpu machine, with 12 cpus - and get the same work done. Vertical scalability is very very hard, it takes decades. If you expect Linux (which scaled 2-4CPUs in v2.4) to scale to 4096 CPUs in v2.6 - and do that well, then you should read more about parallell computing research. Vertical scaling is very hard to do well. To scale, is easy - it is just a matter of compiling Linux to that big machine and run some parallell benches. But to good scaling, is very hard.
                    2.6 is a milestone compared to 2.4 when comes to vertical scaling, the same about some newest 2.6 kernels compared to some older. However, it's hard to say if Linux scales to such numbers of CPUs well or not (according to some SGI notes it scaled well).

                    Not many benches use exactly the same hardware, same software, same compiler settings, etc. Even if the same hardware/software where used, you would argue about compiler settings. Or the branding of the RAM sticks, etc.
                    That's a point and I'm happy you understand this now :>

                    The fact is, Solaris used slower cpu and slower ram, and still topped out. Do you mean that Solaris won, because of the database was so much better than the DB used in Linux benchmark? Is that the reason that Linux only got 87% cpu utilization (which is not that good from a scaling point of view)?
                    I consider database is quite important here.

                    When Phoronix does these benches, where Linux compiles to 64bit binaries with the newest gcc compiler, and OpenSolaris compiles to 32bit binaries with an outdated gcc compiler - you dont complain. Then you think these benches are fair, because Linux wins big time. When you compare Sparc 800MHz to 2.4GHz Intel dual core Linux, then that bench is fair - you think.
                    No, I complained many times before, but then I realized they only benchmark distros defaults.

                    Oh, I must have missed your answer. Can you do a quick recap what Andrew Morton, Linus T, Dave Jones, Alan Cox, etc mean when they say that Linux is buggy, bloated, going to pieces, and code quality is declining? Or is your answer "read the links" - just as usual? Leaving me to figure out, exactly which links, and what you are trying to say?
                    I think you were always taking this out of the contest. They meant only Linux compared to older versions of Linux. This is something natural new bugs and regressions are introduced don't you think? If they say Linux is bloated etc. they don't mean Linux is bloated, more buggy compared to Solaris or any other OS, but to some older Linux' versions.

                    Another question. I dont like FUD or lies. You claim Sun FUDs a lot about Linux. Can you do a quick recap on some of the Sun's FUD here? If Sun really FUDs, then I must criticize that. Normally, Sun plays fair.
                    I mentioned this many times before. Sun anti Linux campaign: Solaris is better Linux then Linux, Bonwick's PR talk about Linux and its community.

                    Comment


                    • #50
                      Originally posted by kraftman View Post
                      I rank Linux over Solaris.
                      You are free to do so. Everyone has the right to an opinion.

                      Originally posted by kraftman View Post
                      It's hard to say which OS is the most stable. It probably cannot be proven...Probably nobody denied Linux has bugs, but you were comparing it to Solaris/OpenSolaris and it sounded like SUN systems are more stable, has less bugs etc. This cannot be proven.
                      You are right it can not be proven which OS is most stable. Some OpenVMS clusters have 100% uptime for 20 years. Single OpenVMS machines can have uptime of 17 years, as I know of. OpenVMS uptime are measured in decades. This is no proof that OpenVMS is more stable than Linux/Unix. But it is an INDICATION that OpenVMS is more stable.

                      Some Unix people say that OpenVMS machines are not patched that often (slow development), so you dont shut them down and therefore you have high uptime, and Unix can easily have similar uptimes if you dont shutdown for patches. There is no stringent proof that OpenVMS is more stable than Unix, but I rank OpenVMS above Unix in terms of stability because I read stories and articles about OpenVMS. I have never read once, that OpenVMS crashed frequently and people said it was a piece of shit, etc etc.

                      Maybe there are so few users so there are very few articles about OpenVMS crashing, and therefore I missed the articles? But I dont think so, everytime I read they say OpenVMS is stable and we know that OpenVMS has a reputation of being stable. And also, where I work, a large finance institution, they say that Unix is not as stable as OpenVMS. There is hear say, articles, blogs, etc - all saying the same thing: OpenVMS is stable. This is NO proof, but an indication: maybe the truth is that OpenVMS is more stable than most OS?

                      If we consider Solaris and Linux. There are many articles, blogs, and hear say, that Linux is unstable. I dont read same thing about Solaris. I read the opposite, that Linux caused trouble when work load increased much and the companies switched to Solaris and everything is fine. I have posted many such stories and articles, I can repost them if you wish. But I dont read stories about big companies switching from Solaris to Linux because Solaris is unstable. Never. I also read stories where people claim Solaris scales better than Linux, but never the opposite. You have seen all these links I have posted. But you have never showed links showing the opposite.

                      You have showed links that Linux exists on big computers and gives high benchmarks on parallell simple workloads. But that is no proof that Linux scales well. You say that Linux exists on as much as 4096 CPU machines.

                      Regarding the Niagara CPU. It is rumoured that Sun's next Niagara CPU: T3, will have 16 cores and 16 threads. It is rumoured that there will be eight T3 in one server. Then it will have 8192 threads. Each thread is treated as a separate CPU by Solaris kernel. Solaris thinks there are 8192 CPUs, there are no practical difference to Solaris between a Niagara thread and a CPU, each thread is individual and therefore you can run 8192 virtual machines (one virtual machine on each thread).

                      When that server is released this year(?), then, because Solaris runs on a 8192 CPU machine (Linux only runs on a 4096 server) - you will admit that Solaris scales better than Linux? Ok, fine. But that is wrong, just because Solaris runs on a server with 8192 CPUs, it doesnt mean that Solaris scales well! Maybe CPU utilization is very low, then Solaris does not scale well. Just because Solaris exist on a 8192 CPU machine, it doesnt prove anything.

                      Fact: I have posted many links that claim that Solaris is more stable, scales better, etc
                      Fact: You have not posted links that show the opposite: that Linux is more stable, scales better, etc. You have showed links that Linux EXIST on big computers, but that is not a proof of scalability.
                      Fact: I have shown links where Solaris SAP scaling is better.
                      Fact: I have shown links where Linux experts say that Linux is good on horizontal scaling.

                      So, even if you dont agree with Solaris is more stable, less buggy, scales better - can we agree that there are far more links/articles/blogs from no Sun people/white papers that say Solaris is more stable, less buggy, scales better, than Linux? I have posted many such links, you have not posted as many. If you google for "Solaris unstable" you get 0.17 million hits, if you google for "Linux unstable" you get 1.8million hits. Can we agree: there are far more links showing that Linux has stability problems, than Solaris has stability problems?

                      Originally posted by kraftman View Post
                      I didn't attack you because of this. I consider you were trolling many times here and that's why I said what I said. Of course I don't want to hide a thing. Some of your trolling started here:

                      http://www.phoronix.com/forums/showp...4&postcount=48
                      Maybe you dont think you attacked me, but I think you attacked me. If you make a joke to a girl, and she thinks that joke was cruel - is it wrong of her to think the joke was inapropriate? Who decides if a joke is cruel or not, you or the victim? If the victim thinks she has been abused, has she been abused - even if you think not?

                      What is ok for you, may not be ok for other people. We are different. You may think you didnt attack me, but I think so.

                      Originally posted by kraftman View Post
                      I don't know what you exactly don't understand.
                      That is my point. You dont understand what I mean, I dont understand what you mean. We think differently. Therefore we must be much clearer, and explain much more. Pretend that I am an idiot, just as you suggested.

                      Originally posted by kraftman View Post
                      It seems it's not documented how many CPUs Solaris can handle running on a single image. However, there are humongous machines which run Linux not Solaris, so this suggest Linux scales better on such machines.
                      No, this is not correct. Let us assume that Linux has 1% CPU utilization, does that mean that Linux scales good on these servers? No. The mere EXISTENCE of Linux on a big machine does not prove nothing.

                      Originally posted by kraftman View Post
                      I provided you links which shows Linux scales up to many CPUs on a single image. You showed me links which only suggest Solaris scales better in some environments, but those aren't evidences (I agree mine also aren't evidences). You said Solaris has reputation of scaling well and Linux has not. It is known SUN was driving anti Linux campaign since years like mentioned here:

                      http://vger.kernel.org/~davem/cgi-bin/blog.cgi/2007/04/10#bonwick_scalability
                      I agree the main architect of ZFS, Bonwick blog, may be anti Linux, but it is not lies nor FUD.

                      Linux v2.4, scaled to 2-4 CPUs according to Linux experts was released 2001. Linux v2.6 which scales to 8-16 sockets according to Linux scalability experts, was released 2003. Two years later. Normal OSes have severe problems scaling to many CPUs, look at Windows. 20 years of development of Windows trying to get into datacenters, and still Windows scales bad even today.

                      The Bonwick blog was written early 2007, just 4 years after release of Linux v2.6. I find it hard to believe that, in four years, Linux went from scaling to 8-16 sockets to thousands of sockets. When that Bonwick wrote that blog, Linux scaled even worse than today. And we see that Linux has problems utilizing as few as 48 cores on SAP benches even today. You must remember that Bonwick did not talk about horizontal scaling, everyone (including Bonwick) knows Linux scales well there. Bonwick talked about vertical scaling.


                      So please, could you do a recap of some of the Sun FUD and lies? I agree that some people (including me) can sometimes be anti Linux when they say how unstable Linux is or how it scales bad, but it is a big difference from FUDing or lying. For instance, Theo de Raadt who is a FreeBSD guy says the same: that Linux is buggy. If many developers (including Linux devs) say the same thing, it is not a proof, but it is an indication it may be correct.
                      Discuss anything Linux related that doesn't fit into any of the other areas.

                      But I am sure that all of us, would choose Linux before Windows. Because Linux and Unix are quite similar.

                      Originally posted by kraftman View Post
                      Those SGI machines were very efficient. There are some factors which can affect CPU utilization. In papers you provided there were different Data Base used and different hardware, so I consider you can't judge scalability just on those papers.
                      How do you know that SGI machines were efficient, in general? As the parallell computing professor said, 15 years ago you could utilize only a few percent of many CPUs. I dont know how the current state of research is, but I bet it is not much better. Everyone says parallell programming is hard to do well.

                      Originally posted by kraftman View Post
                      I've got some different question - does Solaris can handle as many CPUs as Linux? I don't know, but nobody proved if it can.
                      I dont see a reason why Solaris can not handle more CPUs than Linux. Solaris has existed and been shipped on servers with 100s of CPUs, for many years. Even when Linux was in it's infant.

                      Originally posted by kraftman View Post
                      2.6 is a milestone compared to 2.4 when comes to vertical scaling, the same about some newest 2.6 kernels compared to some older. However, it's hard to say if Linux scales to such numbers of CPUs well or not (according to some SGI notes it scaled well).
                      Look at CPU utiliization, to get an indication how well it scales. The SGI notes shows that Linux scales well on parallell work loads, which everyone knows it does.

                      Originally posted by kraftman View Post
                      That's a point and I'm happy you understand this now :>
                      So why does there exist benches if you can not draw any conclusions at all?

                      Comment

                      Working...
                      X