Yes, if you have 229 rules like that, breaking them up is advised. :) The thing is, when you insert objects into working memory, with rules like that you're most likely doing way more criteria matches than you need to. And while I'm pretty sure drools does some rule optimization in the background (like breaking rules like one below into several subrules) it can only go so far. (And the devs will have to address exactly what drools is doing in that regard.) Also, I'm pretty sure drools treats && and "," differently behind the scenes for separating rule conditions. Any reason why your rule doesn't look like this: $bookingRule : BTExecutionBookingRuleModel ( priority == 1, // when a criterion is not set, it is considered as satisfied. (traderCriterion == null|| $dealModel.trader == traderCriterion), (portfolioCriterion == null|| $dealModel.portfolio == portfolioCriterion), // when a product type criterion is set to unknown, it is considered as satisfied whatever deal product type is. (productTypeStringCriterion == null|| productTypeCriterion == ProductType.Unknown || $dealModel.product.productType == productTypeCriterion), (listedIndexCriterion == null|| $dealProductRelatedIndexes.relatedIndexes containslistedIndexCriterion) ) Search the drools-expert docs for "constraintGroup": https://hudson.jboss.org/hudson/job/drools/lastSuccessfulBuild/artifact/t... I believe that behind the scenes, drools creates alpha nodes in the rete based on the constraint groups. So, as your big rules are structured now, you may be getting no benefit from drools' rete optimizations because you have one big group per BTExecutionBookingRuleModel constraint. (Specifically, you may be losing the alpha node sharing, which would cause a big performance hit.) Some other optimization considerations: 1) do your == checks use identity whenever possible? In the case of "productTypeCriterion == ProductType.Unknown" is ProductType.Unknown an enum. (Or, less desirable, an int constant) 2) When you use collections for contains checks, are the collections Sets? (or better yet, EnumSets) ________________________________ From: Costigliola Joel (EXT) <joel.costigliola-ext(a)natixis.com&gt; To: Rules Users List <rules-users(a)lists.jboss.org&gt; Sent: Monday, September 28, 2009 9:57:49 AM Subject: [rules-users] Struggling to improve my rules performance ! Hello, I would like to improve the performance of my Drools process, I have already done some enhancements like removing the use of “collect” and “eval”. I must say I’m a Drools newbie struggling to determine what causes my performance problems. Before asking, I need to explain in few words the problem I want to solve with Drools : My company is a bank where traders are making deals on markets, these deals must be classified in book, this is what we call “booking process”. Booking is done according to booking criteria : which trader has made the deal ? on which product ? etc … A booking rule defines a set of criteria and the target book where the deal will classified. I’m using Drools to : 1. find all booking rules matching deals made by traders 2. apply the right booking rule on the deal among the ones matching the deal (i.e. that is book the deal) I’m testing the booking process with 1000 deals and 229 booking rule. It takes 60 seconds for Drools to book the 1000 deals, which is too slow to be put in production. On the 60 seconds, I have 7 seconds for facts insertion and the remaining for process (after session.fireAllRules) Do you think such numbers are normal/expected ? Like I said I don’t really know where the performance bottlenecks are … I’m thinking of exploring the following possibilities : * splitting my rules into very simple rules * starting the booking process with fewer data and running it several times (instead of one booking process on 1000, run 10 booking processes on 100 deals each) I have put my rules code below (for courageous people), if you see some obvious performance problems or if you can give some hint to explore, you would be very welcome. Thanks in advance for your lights, Regards, Joël Costigliola Ps : rules details I have defined a flow composed of 3 steps/rules group, each group has one rule, here’s the flow and the rules. Flow : ------ "Find matching booking rules" -> "Refresh" -> "Apply booking rule group" Rules : ------ The job of this rule is to store in DealMatchingBookingRules all the booking rule matching a specific deal (each DealMatchingBookingRules instance has a unique Deal). rule"Find matching level 1 booking rules by deal" salience10 no-loop ruleflow-group"Find level 1 matching booking rules by deal group" when $dealMatchingBookingRules : DealMatchingBookingRules($dealModel : deal, $dealProductRelatedIndexes : dealProductRelatedIndexes) $bookingRule : BTExecutionBookingRuleModel ( priority == 1 // when a criterion is not set, it is considered as satisfied. && (traderCriterion == null|| $dealModel.trader == traderCriterion) && (portfolioCriterion == null|| $dealModel.portfolio == portfolioCriterion) // when a product type criterion is set to unknown, it is considered as satisfied whatever deal product type is. && (productTypeStringCriterion == null|| productTypeCriterion == ProductType.Unknown || $dealModel.product.productType == productTypeCriterion) && (listedIndexCriterion == null|| $dealProductRelatedIndexes.relatedIndexes containslistedIndexCriterion) ) then Logger log = LoggerFactory.getLogger("Matching booking rules LOGGER"); log.info("Found a level 1 matching rule for deal : "+ $dealModel + ", rule is "+ $bookingRule); $dealMatchingBookingRules.addMatchingBookingRule($bookingRule); end // used to tell Drools that our $dealMatchingBookingRules are ready for next step in flow rule"Refresh facts in level 1 booking rule process" salience5 no-loop ruleflow-group"Refresh facts in level 1 booking rule process group" when $dealMatchingBookingRules : DealMatchingBookingRules() then update($dealMatchingBookingRules); end rule"Apply level 1 booking rule" no-loop ruleflow-group"Apply level 1 booking rule group" when $dealMatchingBookingRules : DealMatchingBookingRules(hasSingleMatchingLevel1BookingRule == true) then BTExecutionBookingRuleModel effectiveBookingRuleModel = $dealMatchingBookingRules.getSingleMatchingLevel1BookingRule(); effectiveBookingRuleModel.applyRuleOnDeal($dealMatchingBookingRules.getDeal()); retract($dealMatchingBookingRules); end ________________________________ Ce courriel et toutes les pièces jointes sont confidentiels et peuvent être couverts par un privilège ou une protection légale. Il est établi à l’attention exclusive de ses destinataires. Toute utilisation de ce courriel non conforme à sa destination, toute diffusion ou toute publication, totale ou partielle, est interdite, sauf autorisation expresse préalable. Toutes opinions exprimées dans ce courriel ne sauraient nécessairement refléter celle de Natixis, de ses filiales. Elles sont aussi susceptibles de modification sans notification préalable. Si vous recevez ce courriel par erreur, merci de le détruire et d’en avertir immédiatement l’expéditeur. L’Internet ne permettant pas d’assurer l’intégrité de ce courriel, Natixis décline toute responsabilité s’il a été altéré, déformé ou falsifié et chaque destinataire qui utilise ce mode de communication est supposé en accepter les risques. This email and any attachment are confidential and may be legally privileged or otherwise protected from disclosure. It is intended only for the stated addressee(s) and access to it by any other person(s) is unauthorised. Any use, dissemination or disclosure not in accordance with its purpose, either in whole or in part, is prohibited without our prior formal approval. Any opinion expressed in this email may not necessarily reflect the opinion of Natixis, its affiliates. It may also be subject to change without prior notice. If you are not an addressee, you must not disclose, copy, circulate or in any other way use or rely on the information contained in this email. If you have received it in error, please inform us immediately and delete all copies. The Internet can not guarantee the integrity of this email therefore Natixis shall not be liable for the email if altered, changed or falsified and anyone who communicates with us by e-mail is taken to accept these risks. ________________________________