[Bugs] [Bug 1220173] SEEK_HOLE support (optimization)

[bugzilla at redhat.com]

https://bugzilla.redhat.com/show_bug.cgi?id=1220173

Christopher Pereira <kripper at gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Doc Type|Bug Fix                     |Enhancement
           Severity|low                         |high

Red Hat Bugzilla <bugzilla at redhat.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Doc Type|Enhancement                 |Bug Fix



--- Comment #1 from Christopher Pereira <kripper at gmail.com> ---
Besides not using SEEK_HOLE/SEEK_DATA to optimize disk IO, I also discovered
that GB of zeros are send through the network. That hurts.

To reproduce:

- mount a remote gluster volume
- cp a sparse image from the mounted volume to /tmp
- check transfer rates with iptraf

Results:

- holes (GB full of zeros) are transfered via network as raw data

Expected results:

- holes should not be transmited as raw data ("000000000000000...")
- gluster's protocol should send a message like "repeeat 2000 GB of zeros"

Please note that one thing is that gluster client is not supporting SEEK_HOLE,
and other thing is that holes are sent as raw.

Benefits:

- With this optimization, you could copy normal sparse images in seconds
instead of hours!

Moving prio to high, becuase of the big possitive impact for GlusterFS.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
You are the assignee for the bug.