Greetings. We're developing a large rule-based system that will be
used for simulation, and I'm soliciting opinions on an architectural
choice I must make.
The rulebase we're building could potentially grow quite large in
size; a rough estimate might approach 3000 rules. It is also true that
our problem domain can be nicely partitioned into "rule sets"; we
currently do this using a single Rete instance driven with control
facts. Now, it occurred to me that a "better" approach to this
partitioning mechanism would be the instantiation of multiple Rete
instances; each instance would load only those rules belonging to a
particular ruleset, as in:
Rete -- initialization and dispatching rules
* Rete -- equipment ruleset
* Rete -- connection ruleset
* Rete -- alarm ruleset
...
>From a performance perspective (and considering the RETE algorithm as
implemented in Jess), is the latter approach worth doing? It will add
a small bit of complexity to our software, but I can see an obvious
benefit on the maintenance side of the house.
I'd appreciate any comments.
Regards,
--
-------------------------------------------------------------
David E. Young
Fujitsu Network Communications "I claim not to have controlled
([EMAIL PROTECTED]) events, but confess plainly
that events have controlled me."
-- Abraham Lincoln (1864)
"Apology is Policy"
---------------------------------------------------------------------
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]
---------------------------------------------------------------------
