http://bugzilla.spamassassin.org/show_bug.cgi?id=3225
------- Additional Comments From [EMAIL PROTECTED] 2004-04-23 20:50 -------
Subject: Re: RFE: Bayes optimizations
> to either configure for a larger db that uses 32 bit ints or run a
> "super-expiry" more often.
>
> Comments anyone?
My only thought is that without careful or clever design, the super expiry
probably has to hit the entire db, or darn near, and this could result in a
huge spike in disk/processor activity at some arbitrary time.
If there were some reasonable/feasible way to spread that load out over some
amount of time it may be less of a system hit.
I think there might be a reasonable way to do this if the db had two
timestamps representing times N and N+1, where +1 was the current time field
modulo. You could use one bit of a timestamp field to indicate which base
it was relative to. If the actual expiration is < the +1 amount of time,
you should be able to do expiration pretty much at your leasure, as long as
one complete pass through the db could be done in the rollover time period.
Then again, maybe it only takes a few seconds to run the whole super expiry,
and the hit on the db access times and processor utilization aren't worth
considering.
Loren
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.
