Hey Ciao, I want to share my findings while experimenting with Drools 5.5.0.Final and JavaEE 6 on an ARM (embedded) platform: Raspberry Pi and Odroid-U2; and their performances.

I'm going to provide a link below with more details, but to avoid this post being flagged as "shameless plug" :P I make a small summary right here.

I've readapted the waltz benchmark as a JavaEE application / webservice. I'm aware on the debate concerning Expert systems benchmarks, but in this case will be used for a relative comparison of performances. I load-tested the webservice on the Raspberry Pi and the Odroid-U2, comparing the response time to a baseline computer with a 1.8 GHz Intel Core i7. The baseline responds in 132ms. The Raspberry Pi responds in 17219ms (~130x slower) showing it's suffering a lot in this case and possibly is not really a fair use case scenario for the Raspberry Pi, given the intensive computation Vs limits of the JVM on this platform. The Odroid-U2 responds in 566ms (~4x slower) which is a very impressive figures if you consider its specs and cheap cost.

Conclusion I take, in my perspective, there is plenty of potential in using JavaEE on these inexpensive ARM platforms, and I'm looking forward to hear how would be a proper way to benchmark / performance-test of upcoming Drools 6. The reason I'm asking, is because of the following statement on the announcement of Drools 6 on blog.athico.com: "The algorithm addresses the greedy and wasteful nature of Rete. This will make it suitable for more constrained environments, such as mobile devices" - which I do believe is a very similar scenario to the one I've presented above. I'll be happy to clock its performance on the ARM platforms I own, but in this case I prefer be guided what would be a sensible benchmark to highlight the gain given by the new algorithm.

Besides, I hope I provided with this post an informative report to this community :)
Ciao
MM
see also link: http://tarilabs.wordpress.com/2013/08/14/expert-systems-and-javaee-on-arm-a-simple-benchmark/