On Thu, Jan 15, 2009 at 14:19, Michael Alan Dorman <[email protected]> wrote: > Hi, all, > > It took me much longer than I anticipated to get this back-end to > my performance target---comparable to the default DB_File or, when > possible, faster. > > Using the code in masses/bayes-test/benchmark, this back-end runs, > overall, slightly faster than the DB_File code, and 3x faster than > a default MySQL setup. > > Incidentally, if people have some simple tuning changes for MySQL, I'd > be happy to have them so I could produce some more realistic numbers. > I'm running this on an otherwise idle, unoccupied box, so I can tweak > anything that seems appropriate to tweak. > > In the course of development and benchmarking, I've made a number of > changes in that benchmark directory---from updating some config files > so it works with recent versions to rewriting the runmbox.pl script to > use ArchiveIterator instead of Mail::Box and allowing runmulti.pl to > spawn more than just two jobs at a time (to, in theory, show the > impact of concurrent access better). > > I am also trying to come up with a strategy for easily profiling the > whole process using Devel::NYTProf---this has been invaluable in > finding some of my performance issues. > > Anyway, I was thinking that I could submit those while doing a last pass > over the BDB code itself. > > So I am assuming the process for all this is basically the same as for > submitting new rules---add an entry in bugzilla, etc.?
yep! Sounds awesome ;) btw in this case I think it's big enough to need a CLA: http://wiki.apache.org/spamassassin/AboutClas --j.
