Hammer Filesystem update, waiting for FreeBSD 7.0

Hammer filesystem updated, from Matt :

HAMMER continues to progress well.  The blockmap is in good shape   and I am moving on to the next item on the list.  All major functions  work and appear to be stable.

    Three items are needed for an ‘alpha’ release:

    * garbage collection code.  This will be responsible for re-optimizing   the filesystem by scanning the B-Tree and copying data out of partially  free big-blocks, allowing those blocks to be freed.

    * undo fifo (‘instant’ recovery after a crash).  This will handle  all normal crash recovery cases.
      Because data is not overwritten the undo fifo only needs to record meta-data changes such as B-Tree operations, volume header updates,  blockmap updates, and so forth.  Some minor buffer order will also
      be needed to flush the undo data before flushing the actual change, but its a lot easier then ordering requirements for the original design.

    * hard-core recovery code.  This will scan the whole filesystem and guarantee that everything is in a sane state.  Even if it is never supposed to be needed you can’t release a filesystem without a  utility which does this.

    I am not quite calling for testing yet, but please feel free to play around with it with the provisio that there is no crash recovery or  garbage collection code in-place yet.

Meanwhile FreeBSD 7.0 already reach ‘release’ in my testing box, next testing maybe freebsd 8.0 bleeding edge icon smile Hammer Filesystem update, waiting for FreeBSD 7.0


One Comment

  1. Hi Matthew,

    I came across with “Hammer File System” and got impressed with it’s capabilities. I would like to know “How is the Disk I/O performance conpared to Lustre, GlusterFS, ZFS, ext3 and Btrfs. Can it be used as Cloud storage similar to Lustre and Glusterfs where we can combine multiple nodes to have huge storage cloud.

    Also, is “Hammer File System” is going to be available on FreeBSD 8.0 or ant later versions.

    Thank you,

Comments are closed.