These figures are strange indeed, and probably caused by other causes
than simply the loading itself. You should expect a small increase in
load time related to the number of nodes in the process, and some types
of nodes that are using rules will probably take a little longer to load
(as they also require loading rules).
Assuming you are using the same type of splits (rule vs eval) and
dialect (java vs mvel), then loading of a larger process should take a
little longer on average. Not sure how you're producing these numbers
(how to avoid startup cost or gc), but I guess that what you are seeing
is random noise on top of the loading time.
Kris
Quoting Femke De Backere <femmyke(a)gmail.com>:
Hi!
I'm testing the loading time of the line:
kbuilder.add(ResourceFactory.newClassPathResource("nodesflow.rf",
DroolsTest.class), ResourceType.DRF ); for a rule flow file with 1
split node, 3 split nodes, 7 split nodes, ...
But the results I already have are confusing:
1 split node
26071960,8
1
3 split nodes
1018762000
40, 85676594 times the time of 1 split node flow
7 split nodes
4974479,4
1,907968479 times the time of 1 split node flow
Can anybody explain why the 3 split node rule flow file takes a lot
longer than the 7 split rule flow file?
Thx,
Femke
Disclaimer:
http://www.kuleuven.be/cwis/email_disclaimer.htm