[rules-users] Managing memory

[Laird Nelson]

Hello; this is probably a pretty basic kind of question, but I want to see
what other users' tricks are for managing memory in their sessions.

My application is a straightforward survey-type application.  There are
questions, choices, answers (choice selections) and so on.  Once someone
has selected an answer, the rule evaluation kicks in.  The determination of
which question to show next is made based on...well, pretty much anything.
Typically it's on the current user's last answer to a question, but it
might also involve his answers to several previous questions in his own
path through the survey.

This being psychology-related, there are hundreds of questions and probably
hundreds of rules.  "If the question the user just answered is question 4,
and if the user selected 'sometimes' as the answer to question 4, and if he
also selected 'never' as the answer to question 1 during this interview
session, and he is over 40, then show him question 25 next" is a good
example of one such rule.  You can see how there might be nearly one rule
for each question.

(The rulebase session itself could probably be sequential--none of the
facts in the session is modified in such a way that reevaluation of the
rulebase is necessary.)

OK, so, currently I load pretty much everything into the (stateful)
knowledge session I have.  That means--what with all the question text and
answer text and everything--there is a LOT of data in this particular
session.  Multiply times the number of concurrent interviews being
administered and this thing chews through heap in a hurry.

One approach I thought of was to simply load the current question into the
session as a fact and then rely on various accumulate functions to load
other questions and answers as necessary.  But this approach isn't much
better than a linear series of if/then statements, if I understand
correctly.

What are some of the so-called best practices here?  Are there particular
Drools-centric development approaches I should be taking, or architectural
choices I should be making?  Or is this just a case of throw hardware at
the problem?  :-)

Again, I'm not so much looking for A Single Right Answer as I am reports
from the field.  You may feel free to contact me directly if you wish.

Best,
Laird

-- 
http://about.me/lairdnelson
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.jboss.org/pipermail/rules-users/attachments/20111211/fa8de9d0/attachment.html