Announcement

Collapse
No announcement yet.

Benchmarks Of The New ZFS On Linux: EXT4 Wins

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

  • #41
    Originally posted by mgmartin View Post
    I'll preface this by saying I'm a happy production system zfs on linux user and a ( very small ) contributor to the project. ZOL isn't perfect, but many of us use it day to day in production environments with great success. A simple, 1 disk/1 user benchmark doesn't do zol justice. In fact, I think something must have been way off on this article's dbench results.

    I wanted to compare the enterprise performance of zfs vs ext4 on top of mdadm. For time constraints, simplicity, and a comparison to this articles final graph, I only used dbench. The dbench command was always "dbench -t 60 x" where x was 1,20, and 50 simulated clients. All mount options and zfs pool params were left at default values. No l2arc was used. All mdadm / zpool create commands were very basic--no fancy raid offset settings or block size tunings, etc.
    ...
    Thank you! These benchmarks are quite impressive and clearly demonstrate that ZFS can provide excellent performance in real life environments.

    Comment


    • #42
      Originally posted by AndrewDB View Post
      Thank you! These benchmarks are quite impressive and clearly demonstrate that ZFS can provide excellent performance in real life environments.
      I'll just point out that these "real life environments" are all RAID based, which may be typical for a server but not your average desktop.

      Which would line up nicely with where I might recommend ZFS.

      Comment


      • #43
        Originally posted by smitty3268 View Post
        I'll just point out that these "real life environments" are all RAID based, which may be typical for a server but not your average desktop.
        ...
        What exactly is an "average desktop"?

        Obviously ZFS was designed for multi disk, multi TB storage systems, not for a notebook or desktop PC with a single SSD. You have to choose the right tool for the right mission, which is one of the reasons the article that this thread comments upon is flawed (again, as others have already mentioned before, see posts above).

        Comment


        • #44
          Originally posted by smitty3268 View Post
          I'll just point out that these "real life environments" are all RAID based, which may be typical for a server but not your average desktop.

          Which would line up nicely with where I might recommend ZFS.
          which is why you can't buy mainboards with built in raid support. Oh, wait....

          Comment


          • #45
            Originally posted by AndrewDB View Post
            What exactly is an "average desktop"?
            Single disk and no RAID, obviously.

            I agree everyone should use the right tool for the right problem.

            Comment


            • #46
              Originally posted by energyman View Post
              which is why you can't buy mainboards with built in raid support. Oh, wait....
              That said, if you're using motherboard RAID, you're definitely doing it wrong.

              Comment


              • #47
                Originally posted by pdffs View Post
                That said, if you're using motherboard RAID, you're definitely doing it wrong.
                Well yeah... because you're not using ZFS.

                Comment


                • #48
                  Originally posted by smitty3268 View Post
                  Single disk and no RAID, obviously.

                  I agree everyone should use the right tool for the right problem.
                  Actually, the usual "typical desktop" is a Windows PC, so the choice of filesystems is rather limited (i.e. NTFS or... NTFS). If you are using Linux then you are already atypical!

                  For a single disk Linux workstation my personal choice has been reiserfs for as long as I can remember, and I never bothered to really check the performance of reiserfs compared to ext3 or ext4. Nowadays, at the present price levels, for a single disk Linux workstation I would think that choosing an SSD over a hard disk would be much more relevant than choosing any particular filesystem, in terms of performance. For a Linux laptop I would definitely choose an SSD, not only because of the higher speed, but also for the higher reliability (no moving parts) and lower power consumption.

                  In other words, for your "single disk no RAID" "typical desktop" the choice of filesystem is really irrelevant, just like the meaningless pseudo-benchmarking that resulted in the conclusion: "EXT4 Wins".

                  Comment


                  • #49
                    Originally posted by droidhacker View Post
                    First off, this reply should not be taken as representative of my opinions regarding ZFS, but what I really need to suggest to you, is that it is not only ok to support proprietary software, but NECESSARY. It is a very small minded person who thinks that all non-open-source software should be banned in absolute.

                    The problem with your reasoning, is that it IS proprietary software that drives the market. Open source software wouldn't exist if not for proprietary software.

                    The objective, contrary to your opinion, is not to prevent the existence and/or use of proprietary software, but to maintain USER FREEDOM. This is achieved by making the low level aspects of the machine and its control OPEN. An open source OPERATING SYSTEM, web browser, and basic productivity software.

                    The reason why Linux has historically be slow to spread, is because it is generally a hostile environment for proprietary software. For example; Canadian tax day is next week. HOW MANY ACCOUNTING PACKAGES RUN ON LINUX? --- the answer is ZERO. And that is a big problem. Before you try to push people into making their software open, you need them to accept your platform, by making it INVITING to them -- to the developers of proprietary software.

                    As a computer user.... I seek the FREEDOM to run the software I CHOOSE, in an environment that is NOT HOSTILE TO ME. The proprietary operating systems are hostile to ME, so I choose not to use them, but this is a difficult choice, because the only platforms that aren't hostile TO ME, are hostile to the developers of software that I may wish to use. Well, at least in terms of desktop operating systems. Android has bridged that gap for mobile devices, hostile to nobody except apple/ms/rim.
                    Having worked for proprietary software companies for a few years before founding my own company, I have to say no, there is no good reason to support proprietary software, ever. The primary reason for my opinion is simply that projects I worked on have been killed due to neglect from upper management. This is one of the worst possible fates a software developer can ever see for his life's work, it happens all too often in proprietary software companies, and it can't happen in open source.

                    Software freedom is important not only because of the rights that end-users receive. It is important for the rights guaranteed to software developers, and IMO this far outweighs the end-user importance.
                    Last edited by highlandsun; 23 April 2013, 07:35 AM.

                    Comment


                    • #50
                      ZFS is dog slow and eats data what else is new. EXT4 is faster, more reliable and better tested ZFS on Linux is a joke. If you want ZFS stick to Solaris and FreeBSD.

                      Comment

                      Working...
                      X