Paulo J. S. Silva wrote:
Em Sex, 2007-01-19 às 20:03 -0500, Jonathan A. Zdziarski escreveu:
How's that work?
Jonathan
Maybe he means that he is using the Markov stuff to do classification.
Paulo
Yes... CRM114 parsing and Markov calculations. Thank you for that
clarification. I was stating "dspam was configured for CRM114" under the
assumption that if you do configure dspam for CRM114 then several other settings
are clearly not recommended for use (chi-square for example) and admittedly I
further assumed that along with CRM114 comes Markov because that's the
recommended default.
But I believe the operative word here is the CRM114 parsing seems to be on
the order of 30 seconds per message. And I was wondering if this is typical
performance or if I should investigate further into tuning my database.
I'm running an AMD-64 2.? GHz CPU with 256MB RAM and some kind of single disk (I
don't know if it's SCSI/SATA/EIDE...) And I don't have a lot of other
information because it is a virtual machine running Xen.
I realize that 256MB of RAM is not ideal, but that's what they offer for the
price point that I am currently working at. So for the sake of this
conversation let's just say that's all there is going to be right now.
I did find as a comparison that a chi-square classification, which cannot be
used with CRM114 but Bayes only so it's running Bayes, is about 100X faster for
the same physical environment. I've heard rumours of CRM114 performance but was
looking for some anecdotal validation of the actual performance difference that
I am realizing. Perhaps the questions isn't CRM114 but Markov that's the
performance hit. Does anyone else here have any experience with this?
