HAMMER2 design in progress - 1-2 year time frame

Tim Darby t+dfbsd at timdarby.net
Wed May 11 11:36:07 PDT 2011


Just before steam shot out of my ears, I was thinking that this all sounded really cool.  I really like the new concept for PFS functionality.Tim
On Wed, May 11, 2011 at 10:44 AM, Matthew Dillon <dillon at apollo.backplane.com> wrote:

    I'm almost done with the design stage for the HAMMER2 filesystem and
    will soon begin coding.  The basic design document is here:

        http://apollo.backplane.com/DFlyMisc/hammer2.txt

        Lots of tech-speak inside, attach heat sinks to your brain!

    I am going to caution that I expect it to take about a year to implement
    most of the features (including the clustering bits which will be fully
    integrated into the filesystem), and probably 2 years for it to reach
    production stability.  HAMMER2 is not going to replace HAMMER1 any time
    soon.

    In addition, HAMMER2 is going to have two serious restrictions relative
    to other filesystems.  (1) HAMMER2 will not support hardlinks.   And
    (2) HAMMER2 has no physical way to resolve '..' and will depend on the
    operating system's namecache to handle '..' (which DragonFly's will).

    There are many reasons for these restrictions, but it mostly comes down
    to the complexity of cluster cache coherency and mirroring protocols
    (making hardlinks extremely difficult to implement) and support for
    writable snapshots (making inode_num->physical translations and parent
    pointers extremely difficult to implement).  It's kind of a
    one-or-the-other problem.

    HAMMER2 will also do away with the PFS concept, and instead any
    subdirectory tree can be treated as a PFS and independently mirrored,
    and also account for space & inodes used.

    So HAMMER2 is going to have a lot of cluster-friendly features.  A
    veritable ton, but in order to be able to implement those features
    in a reasonable time frame with a reasonably low degree of complexity
    I had to make the above two tradeoffs.

                                        -Matt
                                        Matthew Dillon
                                        <dillon at backplane.com>





More information about the Kernel mailing list