Chris Wedgwood wrote:
Only for metadata, for regular files I think only reiserfs does this.
Doing it probably makes more complicated especially for
writes/flushes.
I think reiserfs unpacks these files when opened in
write for this reason (and optionally packs them again then the file
is closed).
It should not be that tricky for XFS though. Inodes are read in at least
page sized buffer (or 2 page , depending upon the buffer) and are
written back in full buffers In fact, that's how shortform directories
are being handled.
Performance usually suffers so it ends up being a trade-off between
some space-savings and performance. Given disks are insanely large
I'm not sure there is much incentive for this for many people.
If file always remains in local format, the performance should not be
affected (instead it should gain as no extent allocation, no extra I/O
to the disk invariably at a different address). If the format changes,
there will be overhead of copying the inode data to the newly created
block. I am not at all concerned with space saving. I am just thinking
if short form directories can be supported on local inode, what makes it
tough to have the same thing for regular files.
|