Hello,
On one of the drools project I worked before (end if 2011), we were using :
1) Guvnor as rule authoring and repository. Decision tables are updated automatically from remote via the rest api => 100 000 rules in total
2) The package is built on guvnor => 64 bits JVM Heap of 6 GB and on a virtual environment with 2 core and a standard server => 3-5 minutes to build the package
3) the run time are loading the package in binary format => 30 seconds to upload the binary and to build the knowledgebase
4) the project consists of calculating costs during all day, in batches, in real-time, etc.. => 32 bits JVM and in 1.5G of heap, 3 threads are running in parallele and building/removing statefull sessions.
We never came into memory issues but when you run a test scenario, it builds the package so if there are one than one user autoring rules and tests scenarios => Take memory. It i why we decided to go to a 64 bits JVM to be able to have a bigger heap.
I hope it helps
Nicolas Héron
just wanted to ask one of the questions again.
currently, we have around 900 rule assets as shown in Guvnor. some of them are decision tables which mean that the total number of rules might be lot higher ( because each decision table on average has around 20 rows or so)
assuming that the number of rules comes to around 1500 or so,
can we
1. split this one package in Guvnor into 2 packages, build them separately in Guvnor
2. and load each of them one after the other into the knowledgebase?
will this make any difference to the way memory is consumed during loading the rules into the knowledge base?
currently, when loading the single package into the knowledgebase, our windows server is crashing throwing OutOfMemory error.
http://drools.46999.n3.nabble.com/limits-on-number-of-rules-in-Guvnor-in-a-package-loaded-in-knowledge-base-tp3716069p3718794.htmlIf you reply to this email, your message will be added to the discussion below: