11:26 p.m.
Attached is an simple implementation of what I describe below. I was pleasantly surprised
that it ran much faster than I expected, and using less memory. At times in pegged all
four cpu cores on my workstation. It's nice to see such effective utilization of
resources. (Though I did run it with UseParallelGC, so that may have helped.)
The only complaint I have is getting it to stop. :) You've got to kill or ctrl-C to
get it to halt. I am calling drools.halt() in a rule that is being executed. Is there
another way to exit from fireUntilHalt? The console messages printed by the app show that
fireUntilHalt does exit.
I noticed on doing a thread dump that a drools thread is blocking after fireUntilHalt
exits:
Name: Thread-2
State: WAITING on
java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject@4c7e8bf0
Total blocked: 0 Total waited: 414
Stack trace:
sun.misc.Unsafe.park(Native Method)
java.util.concurrent.locks.LockSupport.park(LockSupport.java:158)
java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject.await(AbstractQueuedSynchronizer.java:1925)
java.util.concurrent.LinkedBlockingQueue.take(LinkedBlockingQueue.java:358)
org.drools.concurrent.CommandExecutor.run(CommandExecutor.java:77)
java.lang.Thread.run(Thread.java:619)
CommandExecutor.java line 77 is this:
Command executor = ( Command ) this.queue.take();
"queue" is a BlockingQueue, and BlockingQueue.take() will not exit until it
returns an object from the queue or it's thread is interrupted. Dev d00ds, is there
an official way to interrupt this thread?
--- On Wed, 4/15/09, Greg Barton <greg_barton(a)yahoo.com> wrote:
From: Greg Barton <greg_barton(a)yahoo.com>
Subject: Re: [rules-users] is this kind of App right for DROOLS?
To: "Rules Users List" <rules-users(a)lists.jboss.org>
Date: Wednesday, April 15, 2009, 5:06 PM
A more scalable solution may be this: load all transactions
from site A, start the rule engine in a separate thread
using fireUntilHalt, then start inserting the transactions
from site B. The rules should retract transactions that
match up. (This way ony roughly half of the transactions
are in memory at any given time and you can scale twice as
large.) Once all B transactions have been inserted, what
you have left are all trasnactions that haven't been
able to match. You can save them off and reload them if you
want to attempt matching them in the next round of
processing.
--- On Wed, 4/15/09, Learning BRMS
<rulerhawk(a)gmail.com> wrote:
> From: Learning BRMS <rulerhawk(a)gmail.com>
> Subject: Re: [rules-users] is this kind of App right
for DROOLS?
> To: "Rules Users List"
<rules-users(a)lists.jboss.org>
> Date: Wednesday, April 15, 2009, 4:34 PM
> Uaw.. very nice example, Greg... Thanks
>
> I really like these Drools solution.
> I will try your example with some more heavy class
(with 40
> attributes)
> and expand the volume to 2.6 mi objects coming from a
> hibernate
> session..
>
> I'll let you know.... I think I will enjoy those
tests
> :-D
>
> Thanks again
>
> Kris
>
> On Tue, 2009-04-14 at 23:01 -0700, Greg Barton wrote:
> > Yes, drools can handle this, and can handle the
> volume. You may want to consider sequential mode
execution,
> though. See attached code for some simple matching
rules.
> In the example there's a 5% chance a transaction
with a
> given id will not be present. You could mock this up
> further to have rules to handle matching in the case a
> simple match isn't possible.
> >
> > --- On Tue, 4/14/09, Learning BRMS
> <rulerhawk(a)gmail.com> wrote:
> >
> > > From: Learning BRMS
<rulerhawk(a)gmail.com>
> > > Subject: [rules-users] is this kind of App
right
> for DROOLS?
> > > To: rules-users(a)lists.jboss.org
> > > Date: Tuesday, April 14, 2009, 10:43 PM
> > > Hi people...
> > >
> > > I need to solve a problem on my work and I
would
> hear from
> > > you if Drools
> > > could help us...
> > >
> > > I have two kinds of "connected by
> flatfile"
> > > systems that register our
> > > products selling transactions.
> > >
> > > We need to match dayly every selling
transaction
> against
> > > both systems
> > > and to track any difference for every store.
> > >
> > > An transaction could happens on system A and
not
> and system
> > > B and
> > > vice-versa and on different dates or times
> (connection
> > > problems could
> > > delay the synchronizing).
> > > The transaction mostly have the same ID, but
> there are
> > > manual
> > > transactions too with no id only date and
> values...
> > >
> > > For each situation I have to fire a
different
> action
> > > resolution (today
> > > we have humanhand resolution).
> > >
> > > We should have an 1 million of transactions
by
> day to
> > > process...
> > >
> > > Today we are using a sql script to do
reports
> that will be
> > > humanly
> > > analyzed...
> > >
> > > I want to change this picture as soon as
> possible...
> > >
> > > Could Drools help us with this kind of
analyze,
> and support
> > > this volume
> > > of objects?
> > >
> > > How many objects could we have on working
memory
> at same
> > > time for that
> > > match?
> > >
> > > Thanks a lot for any comments
> > >
> > > Kris
> > >
> > >
_______________________________________________
> > > rules-users mailing list
> > > rules-users(a)lists.jboss.org
> > >
> https://lists.jboss.org/mailman/listinfo/rules-users
> >
> >
> >
> > _______________________________________________
> rules-users mailing list rules-users(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/rules-users
>
> _______________________________________________
> rules-users mailing list
> rules-users(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/rules-users