On Thu, Nov 11, 2010, George Henne <[email protected]> wrote:

>I just checked - PowerMail is responsible for 90% of of the backup
>activity on my system. (I use TimeMachine). The problem is that if just
>one email comes in an hour (which always happens), the complete database
>gets backed up again.

Further on this ... in my instance the database is 860 MB and database
index data (in "Message Database index" package) is 101 MB. Both of
these will change for every new message.

On the other hand, the "Message Database Spotlight cache" package is 571
MB and contains about 107 K items. Most of these are 4 KB which is the
minimum allocation size for my disk.

My TM backup each hour is about 1 GB and takes about one minute.
Admittedly, this is with an internal drive, rather than sending over
Ethernet or wireless so it's not surprising that it's fast.

Trixi's comments, from experience, are instructive ...

On Thu, Nov 11, 2010, Beatrix Willius <[email protected]> wrote:
>
>Database:
>Pro: fast useage, fast copying
>Con: if it's hosed, then it's really kaputt. However, most issues should
>be fixable by deleting the index.
>Incremental backups are not so easy.
>
>Files:
>Pro: simple, problems are usually easy to fix.
>Con: slloooowww searches. Try using AppleMail with a mailbox of 100.000
>mails. This is just a pain.

The work to be done is much the same whether the data is in a database
as we think of it, or the database consisting of files + directory. The
"files + directory" is a database by a different name, after all, just
with different performance characteristics with regard to speed of
access, time to back up, data space allocation on disk, and so on.
There's always a "directory" of some kind, either in the database or the
one in the file system.

The file system has to deal with all kinds of usage patterns so it's
optimized for an average mixture of work. But if you have more
constrained usage then you can improve performance by tailoring for
those, and that's what PM database does. You can, for example, hold a
big chunk of the directory in memory for fast access; something that a
file system won't allow you to do directly. I could go on (a lot) but
you get the point - the PM database is great for storage and access
efficiency, but not so good for incremental backup.

I'd like to see an installation option for PM to place its database on a
sparse-bundle-disk-image (for those users on Leopard and later). All the
"plumbing" for this already exists (you could do it yourself if you
wished) so it's not a big development effort. But it would be non-
trivial for testing and QA (I'm by no means saying that the total effort
is trivial, just that Apple already provides the support tools).

Regards.....Peter


Reply via email to