[Kde-pim] Akonadi: single database design mistake?

[Dmitry Torokhov]

On Tue, Nov 29, 2011 at 09:49:57PM +0200, Andras Mantia wrote:
> Dmitry Torokhov wrote:
> 
> > so any query has to traverse entire index for all folders in all
> > accounts one might have.
> 
> Don't forget this is a database server, we use (database) indexes, so - in 
> theory - this might be just as fast or even faster than doing what we did in 
> KMail with the per folder indexes.

The fact that it is a database does not make it magically work
significantly better. Database is good on dedicated server with
multitude of concurrent users that often query similar data. Sich setup
will have large caches which are hot and thus serve users better. Here
we have mostly single user scenarios (yes, ther are  background tasks
but they are few and far between) so aside of standard API database does
not provide many benefits.

Again, the difference that before there was an index per mailbox folder and as
long as folder was of reasonable size (under 10K mails I'd think)
reading new index into memory was very quick and then it stays in buffer
cache and all is well. Single database setup forces us to access indices
that span all folders in all accounts.

Thanks.

-- 
Dmitry
_______________________________________________
KDE PIM mailing list kde-pim at kde.org
https://mail.kde.org/mailman/listinfo/kde-pim
KDE PIM home page at http://pim.kde.org/