Memory usage is -very- dependent on the nature of the rules and the
size of the fact-list. Writing rules to minimize memory use is
conceptually simple: think of each rule as a laundry list. What Jess
does is collect all the 'partial matches' of facts to the lists of
patterns on rule LHSs. It starts from the top of a rule and works
down. Your job is to write rules such that Jess will generate the
smallest number of half-completed lists. Put rarely-matched patterns
near the top of the rules. Put the most specific patterns at the top
of the rules. These are the general guidelines. SOme experimentation
is generaly a good idea.
I think Duane Morin wrote:
>
> Ok, the system I've built is thus:
> a fact "sentence" is asserted that contains a sentence (for example,
> "tell me about 401k plans")
> rules fire
> as a result, facts are asserted that refer to product-specific concepts
> that I have been able to extract from the sentence (in this case,
> (match (concept KEYWORD)(type RETIREMENT-PLAN))
> (match (concept PRODUCT)(type 401K))
>
> I then use regular Java to grab these match facts and act accordingly.
>
> Well, this is a web-based service. So many people are gonna hit it at once.
> I'm currently looking into how to best implement the engine so that I
> don't just have n copies of it running around.
>
> Can someone give me a brief lesson (or tell me where I can get one) on
> exactly what sort of memory performance Jess has? If I have 3 rete
> engines running around my memory space, is taht going to decimate it?
> What if I had 30 engines?
>
> For the record, we're talking about what I think is a relatively small
> ruleset - less than 200 rules (probably ALOT less). Also, the number
> of simultaneous users will probably not exceed a dozen. So what I'm
> thinking about is setting up a connection pool to a handful of engines
> (like maybe 3-5) and if it's necessary, just queue people's requests to
> use them. Part of this, too, is gonna be how long a request takes to
> finish, and I don't have that info yet. Right now it takes like 5 seconds,
> but that's on a slow box with non-optimized code.
>
> Duane
>
>
> ---------------------------------------------------------------------
> To unsubscribe, send the words 'unsubscribe jess-users [EMAIL PROTECTED]'
> in the BODY of a message to [EMAIL PROTECTED], NOT to the
> list. List problems? Notify [EMAIL PROTECTED]
> ---------------------------------------------------------------------
>
>
---------------------------------------------------------
Ernest Friedman-Hill
Distributed Systems Research Phone: (510) 294-2154
Sandia National Labs FAX: (510) 294-2234
Org. 8920, MS 9214 [EMAIL PROTECTED]
PO Box 969 http://herzberg.ca.sandia.gov
Livermore, CA 94550
---------------------------------------------------------------------
To unsubscribe, send the words 'unsubscribe jess-users [EMAIL PROTECTED]'
in the BODY of a message to [EMAIL PROTECTED], NOT to the
list. List problems? Notify [EMAIL PROTECTED]
---------------------------------------------------------------------
