If the release date of M1 is not to far I can surely re-execute the test against that release. For my evaluation: do you have a roadmap/release schedule for JBoss Rules 3.2? (We don't want to run a production system on beta or unreleased code) Thanks a lot for your answer :) ________________________________ Da: rules-users-bounces(a)lists.jboss.org [mailto:rules-users-bounces@lists.jboss.org] Per conto di Mark Proctor Inviato: venerdì 26 gennaio 2007 17.09 A: Rules Users List Oggetto: Re: [rules-users] Performance of JBoss Rules Dandrea, I believe that MVEL has just got stable for JDK1.4 so we should be doing an M1 release which is much much much faster than 3.0.x - it should be backards compatible, any chance you could re-run those tests when M1 is out? Thanks Mark Dandrea Francesco wrote: Hi all, I have to evaluate if JBoss Rules is a viable solution in this scenario: We have to monitor a large network using performance alarms. These alarms arrive every 15 minutes. We don't know the real rate, but peaks of many tens of thousands of alarms every 15 minutes must be managed in a timely way. We want to use JBoss Rules to correlate these alarms, in order to show less higher level alarms to the users . We have basically 2 use cases: 1) I have to correlate the alarms on a single node of the network. (for example in node A I have 2 alarms of different severity about the connection with node B, so I aggregate everything in a third alarm saying "I cannot connect to node B from node A"). In this scenario I have many workingMemories (one for each node) and not so many facts for each working memory. The performance of JBossRules is very very good (almost linear with the number of nodes) 2) I have to correlate alarms among the nodes. (for example all nodes connecting with a node A have an alarm "I cannot connect to node A from node ...", I want to create an alarm "The node A is unreachable"). In this scenario I'm very worried about the performance of JBoss Rules, as I prepared a test case (in attachment) and I inferred that the growth is highly non linear on the number of the facts. Is this performance data wrong? Am I doing something stupid? Can someone comment? Here are the numbers (The numbers are not so important, but rate they grows is) Number of Facts Time to assert alla the new allarms (seconds) Time to modify the alarms to close them (seconds) Total 1000 1.735 0.969 2.704 2000 5.875 3.000 8.875 3000 12.281 6.406 18.687 4000 23.672 11.672 35.344 6000 52.282 23.250 75.532 8000 87.188 42.532 129.720 12000 200.767 94.642 295.409 16000 360.909 176.579 537.488 20000 518.019 263.158 781.177 Or in a graph: Thanks a lot for your interest. -------------------------------------------------------------------- CONFIDENTIALITY NOTICE This message and its attachments are addressed solely to the persons above and may contain confidential information. If you have received the message in error, be informed that any use of the content hereof is prohibited. Please return it immediately to the sender and delete the message. Should you have any questions, please contact us by replying to webmaster(a)telecomitalia.it <mailto:webmaster@telecomitalia.it> . Thank you www.telecomitalia.it <http://www.telecomitalia.it> -------------------------------------------------------------------- ________________________________ _______________________________________________ rules-users mailing list rules-users(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/rules-users -------------------------------------------------------------------- CONFIDENTIALITY NOTICE This message and its attachments are addressed solely to the persons above and may contain confidential information. If you have received the message in error, be informed that any use of the content hereof is prohibited. Please return it immediately to the sender and delete the message. Should you have any questions, please contact us by replying to webmaster(a)telecomitalia.it. Thank you www.telecomitalia.it --------------------------------------------------------------------