Announcement

Collapse
No announcement yet.

FreeBSD 8.0 Benchmarked Against Linux, OpenSolaris

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

  • #16
    Originally posted by yogi_berra View Post
    @ kraftman - What makes you think Oracle will change anything?
    I saw some interesting article at lwn :>

    Comment


    • #17
      Again, can someone explain how to make OpenSolaris kernel run in 32bit mode? Last time, if I remember corectly, Freebsd won more benchmarks against OpenSolaris. If Solaris kernel ran in 64bit mode here it would be fun :>

      @Kebbabert

      You never give up :P

      Btw. You know the answer, don't you?
      Last edited by kraftman; 12-02-2009, 07:16 AM.

      Comment


      • #18
        FreeBSD 8 0 Benchmarked Against Linux OpenSolaris

        The Linux NTFS project seems pretty stable as far as reading NTFS volumes, but write support is very limited You can change data of a file, but arent able to extend files or create new ones. There is some support for dynamic volumes, but again, its not "production" software.

        What exactly is your reason for moving to Linux anyway?

        Comment


        • #19
          Another item is that if they do default installs on single partitions on BSD then UFS normally doesn't install with softupdates enabled. Any "sane" admin would install a *nix box with multiple partitions for /, /tmp, /usr, /var, /home and possibly others anyway. Softupdates should speedup the file read/write tests and possibly others as well.

          And no, turning softupdates on for the / is not what you want. You want your / to be the most stable in case the other filesystems have issues, so you can still boot to single user mode and recover them.
          Last edited by rhavenn; 12-02-2009, 12:45 PM.

          Comment


          • #20
            Originally posted by kraftman View Post
            Again, can someone explain how to make OpenSolaris kernel run in 32bit mode? Last time, if I remember corectly, Freebsd won more benchmarks against OpenSolaris. If Solaris kernel ran in 64bit mode here it would be fun :>

            @Kebbabert

            You never give up :P

            Btw. You know the answer, don't you?
            I dont know how to make OpenSolaris run in 32bit mode. It chooses the correct version upon boot. If the cpu is 64 bit, then it boots in 64 bit mode. Solaris 32 bit version is not really good, Ive heard. It has been 64bit since ages, a long time ago. If the CPU is 64 bit, then Solaris should have booted in 64bit mode. Remember, there are not different Solaris versions of the kernel, depending on if the machine has many many CPUs, or few, or if it is a supercomputer or a netbook, or if it is 32bit or 64bit. It is the same install CD and the same kernel. This is true scalability. It scales from small netbooks up to whatever you wish, with the same kernel. No need to recompile or redesign or strip out code.



            Regarding OpenSolaris winning benchmarks or not. As I told you, OpenSolaris is under heavy development and the focus is not performance, as of now. The focus is functionality, not stability, nor performance. Developers just add all sorts of functionality right now. And very often the functionality is buggy, this is true. Read the OpenSolaris forums, and you will see. It is buggy because it is alfa stage. Not even Beta. Things change pretty fast. In one build, opensolaris performance may suck, in the next build it is very very fast. And the build after that, the functionality is broken and will not work! This has happens all the time, and it is a pain. You have choose which build you want to use. This is true. Read the opensolaris forums and you will see I speak true. Some builds are buggy, others are better. This is because every other week, the source is checked out and a OpenSolaris build is released. Of course such builds are buggy, they are straight from the source development tree! But can compete with other OSes, still. Testament to good code quality. Try to check out code from another OS under construction, and build it, and try to bench it, see how stable/fast it is. And compare that to OpenSolaris builds. OpenSolaris will win hands down.

            In my opinion, it is not really fair to compare something under construction to products that have left the beta stage. It is like comparing the FPS of a game under construction vs a fully patched and released game - not really fair? Just before release, the game developers switches focus to performance and raises the FPS a lot. Isn't this true? The same here. No one can expect a product in alfa stage to be optimized for speed.

            Therefore, I dont really care of OpenSolaris performance right now. When it is released, I will care. Right now, performance changes from build to build (which all these Phoronix benches show, sometimes OpenSolaris gains performance, sometimes it loose). But OpenSolaris has the ability to become very fast when it is released later, and left this alfa stage.
            Last edited by kebabbert; 12-02-2009, 02:37 PM.

            Comment


            • #21
              Originally posted by kebabbert View Post
              I dont know how to make OpenSolaris run in 32bit mode. It chooses the correct version upon boot. If the cpu is 64 bit, then it boots in 64 bit mode.
              Ok, thanks.

              Comment


              • #22
                Booting into 32-bit mode on a 64-bit proc can be done as so listed here:

                http://blogs.sun.com/alta/entry/boot_into_32_bit_kernel

                Comment


                • #23
                  Originally posted by deanjo View Post
                  Booting into 32-bit mode on a 64-bit proc can be done as so listed here:

                  http://blogs.sun.com/alta/entry/boot_into_32_bit_kernel
                  Ok, thank you Deanjo.

                  Comment


                  • #24
                    Originally posted by kraftman View Post
                    Ok, thank you Deanjo.
                    No problem, it just happened to be in one of my rarely used bookmark folders (the one called solaris).

                    Comment


                    • #25
                      hi,

                      Can author of the article confirm that Solaris was running in 32bit ?

                      OpenSSL RSA 4096-bit test is a bit suspect!

                      -thanks-

                      Comment


                      • #26
                        Originally posted by xosevp View Post
                        hi,

                        Can author of the article confirm that Solaris was running in 32bit ?

                        OpenSSL RSA 4096-bit test is a bit suspect!

                        -thanks-
                        These results should be expected as the OpenSSL that comes with OpenSolaris uses the Crypto Framework.

                        Comment


                        • #27
                          Here is a SAP benchmark on 32 cores, which is more interesting for benching scalability and large scale systems for Enterprises. Solaris 10 is 25% faster than Linux, using slower CPUs. Remember, this is not OpenSolaris which is in alpha stage. But it is S10 which is sharp released OS. However, S10 will be replaced with modern OpenSolaris. It will kill everything, when done.

                          Solaris: 8 CPU, 32 cores. 2.6 GHz
                          Linux: 8 CPU, 32 cores. 2.8 GHz
                          Windows: same setup as Linux

                          http://blogs.sun.com/SAPonSun/entry/10_000_saps_do_more
                          See the official pdf benchmark from SAP website here.

                          The problem is utilization. Solaris is close to 100% utilization on all cores. Windows is down to 87% utilization on all cores. In other words: Windows doesnt scale to well. Nor does Linux. The cores are not fully pegged. Solaris cores are fully pegged so scales better: 25% faster than Linux, on slower hardware. 21% faster than Windows, on slower hw.

                          Comment


                          • #28
                            Originally posted by drdoug007 View Post
                            These results should be expected as the OpenSSL that comes with OpenSolaris uses the Crypto Framework.
                            Such results shouldn't be expected, because Linux and *BSD also have this.

                            @Kebbabert

                            I believe nobody's interested in Suns propaganda. You say Windows scale better then Linux... However, you proved few times you're a troll, so your posts can be safely ignored. Second time on an ingore list.
                            Last edited by kraftman; 12-15-2009, 06:09 AM.

                            Comment


                            • #29
                              Originally posted by kraftman View Post
                              @Kebbabert

                              I believe nobody's interested in Suns propaganda.
                              But SAP has validated and approved those benchmarks. The benchmarks are checked and certified by SAP. On the "SUN propaganda web page" there are links further to SAPs home page showing those certifications. I copy & paste those links for you, because you do not believe links on SUN web page. Here are those links, they are not on SUN web page anymore, those links are now free and "cleansed" from SUN propaganda:

                              Windows certified SAP result on 32 cores, from SAP home page:
                              http://download.sap.com/download.epd...0790148C9F4C1E

                              Linux certified SAP result on 32 cores, from SAP home page:
                              http://download.sap.com/download.epd...FCA652F4AD1B4C
                              We see that Linux only achieves 87% utilization on the cores. This shows that Linux does not handle many cores well, on a single machine.

                              Solaris certified SAP result on 32 cores, from SAP home page:
                              http://download.sap.com/download.epd...11DE75E0922A14

                              I understand that if SUN links to those white papers, you guys dismiss those white papers as "propaganda". But if SAP hosts those white papers, then it is not SUN propaganda anymore. Then it is maybe SAP propaganda.

                              Or, maybe SUN should not have pointed to those links. Instead SUN should have waited until you guys discovered them by yourself. If you discover that white paper by yourself, then you could not accuse them of being SUN propaganda. If SUN links, then that link is useless propaganda. The trick is that SUN must not link. But how will we discover that link, if no one points them out?



                              In hind sight, I should have pointed to those white papers on SAP web page myself. Then you guys could not have accused them of being SUN propaganda. Next time, I will not point to SUN web page (that points to the benchmark/white paper). Instead, I will point to the white paper/benchmark directly myself.
                              Last edited by kebabbert; 12-14-2009, 08:29 AM.

                              Comment


                              • #30
                                Oops, my bad. The CPUs are six core, not quad core. So all machines have 8 CPUs and 48 cores. Which is clearly stated if you closer read the white paper on SAP home page.

                                Comment

                                Working...
                                X