Hey Mathew:
Hope you are having a great weekend.
You make a good point on benchmarking. I suspect defining the criteria model
is at issue.
I rely on some experience building search engines to say for
high-connection-rate 'small results' there is no level where anything but a
flat file is faster than MySQL. Clearly DbMail 's original planners made
thoughtful and wise choices with the critical mission in mind.
Certainly compilation of complex data models is best left to a true object
relational database Management System (DBMS) like PostgreSQL and others,
but this is email storage and retrieval we are talking about, not a million
sensor-point fluctuation and load analysis system for the European
interstate power grid.
Perhaps where an object relational database DBMS or another quasi (MySQL) or
full-standards relational database management system might perform better is
on a high volume of connections each returning a very large result. (I don't
refer to performance as being a measure of how difficult coders find queries
are to create but what is the value for the end user.)
Forget features like Triggers and stored procedures for a moment and
consider the criticial mission of fast email storage and retrieval. What
might interfere with this goal over time may be less about the growth in
mail box sizing but perhaps more about the increase in indivdual message
block/object size (i.e.: consumer digi-cam and music files).
I don't know what would be the elevated data-flow burden ('y' connection
rate @ 'x' result size) where PostgreSQL or anything, if it is true, becomes
faster than MySQL for dbmail in terms of mail blocks size average on
connections per minute. Perhaps someone has experience over conjecture on
that question. In other words, at what connection rate 'y' with 'x' average
result size does anything outperform the currently popular DbMail/MySQL
combo? Thereafter, what fact analysis-based dbmail/database recommendation
could be made to a mail system planner?
... just a thought.
Have many happy days...
best... Mike
----- Original Message -----
From: "Matthew" <[email protected]>
To: "DBMAIL Developers Mailinglist" <[email protected]>
Sent: Sunday, October 10, 2004 8:51 AM
Subject: Re: [Dbmail-dev] Re: Gerrit's Speed issue............
> Aaron Stone wrote:
>
> >So... all criticisms of MySQL are exactly on target. But so what? On the
> >down side, we need to eschew some features and best practices in order to
> >support MySQL. On the up side, support is actively being added to MySQL
> >and we can just roll right along with it, incorportating features into
> >future version of DBMail and raising the bar on our minimum MySQL
> >requirements as we go.
> >
> >
> >
> At this point I know we are not going to drop mysql, but it just
> frustrates me that we have to spend 1 second of time working around
> limitations. But enough on that topic :-)
>
> >It would be really, really nice if someone could do some latest-version
> >benchmarks between MySQL and PostgreSQL. Bouncing around heresay and
> >personal preferences only makes me want to support both camps even more
> >:-P
> >
>
> I couldn't agree more. Some recent updated benchmarks between mysql and
> pgsql in general would be interesting, but more interesting to this
> mailing list would be some dbmail benchmarks comparing pgsql and mysql.
> My guess as to the results would be that mysql is a little faster when
> only a few users are connected. I also think that pgsql could be much
> faster all around if the code was developed to take advantage of some of
> it's features.
>
> Matthew
>
> _______________________________________________
> Dbmail-dev mailing list
> [email protected]
> http://twister.fastxs.net/mailman/listinfo/dbmail-dev
>
