On Mon, 2008-08-11 at 14:37 -0700, Hyrum K. Wright wrote:
> A couple of months ago, I started experimenting with storing Berkeley
> db deltas in the same way we store FSFS deltas. Instead of keeping a
> fulltext at HEAD and redeltifying previous revisions against it, this
> new method stores one fulltext at the first revision and then builds
> deltas against that, using our typical skip deltas algorithm. The
> space savings are pretty dramatic
Huh. I wonder if BDB is failing to garbage-collect the space used after
a rep is redeltified?
On a related note, it would be interesting to implement an optional
representation cache alongside the FS, to ease the runtime penalty of
this approach in both BDB and FSFS where it's already used. Such a
cache has security and deployability issues when the FS might be
accessed by multiple uids, but should be a win in the typical server
configuration.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe_at_subversion.tigris.org
For additional commands, e-mail: dev-help_at_subversion.tigris.org
Received on 2008-08-12 01:15:45 CEST