Michael, > We had some issues months ago with DCC 'hanging' I dccproc (consumed all > available ram and cpu) and with the change in dcc licensing, we disabled > dcc. Since we are using amaivsd-new and some of the 'rulesdejour' (sa has > 700 rules, we added 2000 more) we have not seen a measurable difference in > accuracy.
> Has anyone else done a study to see if dropping dcc actually > affected the accuracy of the scores? > ... if you think the additional cpu and ram cycles are worth > even pursuing licensing their patent of the derived GPL code. I haven't done a study on how DCC affects accuracy, but I did a performance benchmarking comparing DCC/Razor2/Pyzor/RBL (measuring throughput), and I have some anecdotal evidence that DCC often added the few decisive score points that saved us from certain spam where other tests apart from Bayes mostly generated a OK. As far as comparing the CPU load, the DCC (even dccproc) is the most lightweight (apart from RBL and other DNS tests), then comes Razor2 (but has the longest latency), and finally Pyzor, which is the most heavy CPU consumer. Mark ------------------------------------------------------- SF.Net email is Sponsored by the Better Software Conference & EXPO September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf _______________________________________________ AMaViS-user mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/amavis-user AMaViS-FAQ:http://www.amavis.org/amavis-faq.php3 AMaViS-HowTos:http://www.amavis.org/howto/
