[Gluster-users] Usage Case: just not getting the performance I was hoping for

[Jeff Darcy]

On 03/15/2012 09:20 AM, Brian Candler wrote:
> I pointed out the problem on the list, and was told that no one strategy can
> fit every usage case.  This one is particularly tuned to Samba (which writes
> 128K past the end of the file then back-fills with data).  It was suggested
> that glusterfsd could avoid this problem by using ftruncate before writing.

Unfortunately, that approach would mean that we're constantly preallocating and
then deallocating space with every write.  The approaches under consideration are:

* Set allocsize=0 when creating XFS filesystems if sparse files are expected,
to turn off preallocation altogether.  This can be done today, at some cost for
files that are actually contiguous.

* Fix the pathologically over-optimistic preallocation heuristic in XFS.

* Modify the stripe translator so that it "collapses" sparse stripes into
larger contiguous regions.

These evaluations are still ongoing.  We'll let folks know when we reach some
conclusions.