On Fri, Jan 16, 2004 at 10:44:04AM -0500, Damian Gerow wrote: > Thus spake Ed Walker ([EMAIL PROTECTED]) [15/01/04 20:04]: > > > > How's this holding up under heavy load with large clusters? How much > > overhead for either the SQL server or SA? > > Having not written the code, and not being in a position to test it, I > couldn't say for sure. But perhaps Larry Rosenman can comment: > > <http://bugzilla.spamassassin.org/show_bug.cgi?id=195> > > IIRC, he's previously stated that updating the database is slower, but > checking the database is faster. Correct me if I'm wrong...
I can't comment on how well it handles load simply because I don't have a huge load, but it should scale as well as your database scales. Indeed, the updates are what makes the process slow. The Bayes SQL code does away with the concept of a journal (since the goal was to not involve the filesystem at all). So, selects are very fast but each used token has to be updated which slows things down. I'm currently working on another module that provides an in database journal that hopefully will speed things up a bit. Also, an RPC based module that makes calls to a seperate RPC bayes server that will allow for some DB optimizations and caching that we can't do in spamd. Michael ------------------------------------------------------- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
