On Mon, Dec 10, 2012 at 06:05:59PM +0000, Steven Whitehouse wrote:
>
> If the block device can zero out blocks more efficiently than just
> writing zeros (i.e. via sb_issue_zeroout()) then this could be faster.
> In fact GFS2 already zeros out fallocate()d extents in this way because
> it has no way to mark unwritten extents in its metadata, and we did not
> want to leave them uninitialised,
Sure, but if the block device supports WRITE_SAME or persistent
discard, then rpesumably fallocate() should do this automatically all
the time, and not require a flag to request this behavior. That is, a
seek plus writing 1MB does take more time than the amount of disk time
fraction that it consumes if you compare it to a seek plus writing 4k
or 32k. It may be that 32k is too low, and there may be some
disagreement about whether people want to the system to automatically
tune for average throughput vs 99.9 percentile latency.
Regardless, this is actually something which I think the file system
should try to do automatically if at all possible, via some kind of
auto-tuning hueristic, instead of using an explicit fallocate(2) flag.
(See, I don't propose using a new fallocate flag for everything. :-)
- Ted
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
