> >3) Find out how much free memory you have without spamd running. ... > >4) Divide the free memory by your answer from 2. That should give you a > > good rough-estimate number to work with.
> As an alternative to the above, you can calculate an approximate upper > limit for "m" by taking the resulting memory size calculated in step 2 and > dividing the total memory in the system by the step 2 number and rounding > down i.e. if the ... While it is true that the amount of memory determines the upper bound on the number of running SA processes, it is not the only limitation, remember that SA is CPU intensive. One should not be misled to believe that adding lots of memory will allow to run arbitrary many spamd (or amavisd-new) processes. See http://www.ijs.si/software/amavisd/amavisd-new-magdeburg-20050519.pdf for some of my benchmarks. The $max_servers (or maxproc) setting in amavisd-new (Postfix) is pretty much equivalent to -m on spamd, so the results should mostly apply to spamd as well. In particular, see slides 10, 11, and others (like 21, 23, 25, 26). In each case the y axis show total system througput (MTA + content filter) in messages per second, and x-axis shows a number of concurrent content filtering processes. Mark
