Re: [rules-users] Showstopper issues for Drools Adoption in our organization
Kapokfly, For issue 1: I have a very similar issue. What I am currently doing, is to pre-load my reference data from DB into my JVM, but I do not load those entities into working memory. Instead, I have the rules written to use the FROM clause in the left hand side of the rule. Example: When Customer ( $indCode : clientIndustry, activeSw == true ) $industry : Industry () from lookupIndustry( $indCode ) What this does, is obtain the client Industry code from my customer object, and capture the value to the $indCode variable. I then capture all Industry objects into the $industry variable that are returned from the call to the function lookupIndustry(). The lookupIndustry function is purpose built to locate the specific industry objects that meet the query criteria provided as the function argument. In this case, I am looking up the industry from a map of Industries, with the key of the map set to the unique Industry code. In my real-world scenario, I would prefer to rely on a no:sql DB to perform the lookups for me. This would buy me the performance benefits of my key:value maps being driven by faster C/C++ code, while removing the need to me to code such lookup and caching services in my own code. In addition, my caching process, does actually cache my DB objects in the JVM memory, but instead it preloads my no:sql DB with all my facts, and using a Real Time Messaging system, I intercept all DB updates to sync up my no:sql data. The no:sql is not always the best fit, but because of the types of lookups I need to perform, no:sql can retrieve my facts from DB and return them to my code as instantiated objects much faster can JPA, or native SQL queries against my RDBMS, and with much less memory impact than locally coded JAVA data maps. Two million records is a lot, regardless of the actual data content, so it seems to me you might want to either query SQL using an function, or querying a no:sql system (like mongodb) and then using this logic to process on the data in almost the same way you would had you loaded the data in working memory of the knowledge session. Hope this helps. Regards, Armand Welsh -Original Message- From: rules-users-boun...@lists.jboss.org [mailto:rules-users-boun...@lists.jboss.org] On Behalf Of kapokfly Sent: Friday, November 11, 2011 2:54 AM To: rules-users@lists.jboss.org Subject: [rules-users] Showstopper issues for Drools Adoption in our organization Not sure if these are also your questions you might have with the rule engine adoption, but we have seen 2 showstopper issues for drools adoption in our organization, or it is just because we are lack of some necessary knowledge. Issue 1) : pattern match for large data There are cases in a rule we need find matched records against a large population of data and perform some actions. Example: From 2 millions employee directory, find those belongs to Depart1 and with salary increased by 10% last year. Inserting all these employee's data into WorkingMemory is HUGE and mission impossible, we can't simply load the entire db into the working memory, do we have any other approach to accomplish or this is just not a use case can be resolved by a rule engine? Issue 2): Map business object to real engineering implementation What we are trying to do is to build a rule UI which a business user without engineering background of our codebase can work with; the business object we show on the UI might be different with the real engineering objects already in system, for example, on the UI a user might be able to able to see something like person.address.addressLine1 but the actually implementation might not be an object traverse implementation. Example, the model could be Long Person.addressId ; then referring to his address we need another lookup and we want to hide such complexity to the end user. The approach we took is to design a translation layer to translate the user input to drools rule string, is this the right approach? If not, any suggestions? I know Drools is mainly designed for developers but just trying to see if there is any way we can work it around. Thanks and looking forward to your recommendations. Ivan - Ivan, your Panda, forever -- View this message in context: http://drools.46999.n3.nabble.com/Showstopper-issues-for-Drools-Adoption-in-our-organization-tp3499367p3499367.html Sent from the Drools: User forum mailing list archive at Nabble.com. ___ rules-users mailing list rules-users@lists.jboss.org https://lists.jboss.org/mailman/listinfo/rules-users ___ rules-users mailing list rules-users@lists.jboss.org https://lists.jboss.org/mailman/listinfo/rules-users
[rules-users] Showstopper issues for Drools Adoption in our organization
Not sure if these are also your questions you might have with the rule engine adoption, but we have seen 2 showstopper issues for drools adoption in our organization, or it is just because we are lack of some necessary knowledge. Issue 1) : pattern match for large data There are cases in a rule we need find matched records against a large population of data and perform some actions. Example: From 2 millions employee directory, find those belongs to Depart1 and with salary increased by 10% last year. Inserting all these employee's data into WorkingMemory is HUGE and mission impossible, we can't simply load the entire db into the working memory, do we have any other approach to accomplish or this is just not a use case can be resolved by a rule engine? Issue 2): Map business object to real engineering implementation What we are trying to do is to build a rule UI which a business user without engineering background of our codebase can work with; the business object we show on the UI might be different with the real engineering objects already in system, for example, on the UI a user might be able to able to see something like person.address.addressLine1 but the actually implementation might not be an object traverse implementation. Example, the model could be Long Person.addressId ; then referring to his address we need another lookup and we want to hide such complexity to the end user. The approach we took is to design a translation layer to translate the user input to drools rule string, is this the right approach? If not, any suggestions? I know Drools is mainly designed for developers but just trying to see if there is any way we can work it around. Thanks and looking forward to your recommendations. Ivan - Ivan, your Panda, forever -- View this message in context: http://drools.46999.n3.nabble.com/Showstopper-issues-for-Drools-Adoption-in-our-organization-tp3499367p3499367.html Sent from the Drools: User forum mailing list archive at Nabble.com. ___ rules-users mailing list rules-users@lists.jboss.org https://lists.jboss.org/mailman/listinfo/rules-users
Re: [rules-users] Showstopper issues for Drools Adoption in our organization
Ivan, The more complex your problem gets, more technologies you will need to employ to solve it properly. Indeed Drools does not work directly with persistent storages, because its focus is on the efficient processing of business rules, processes and events, delegating the management of persistent and historical data to specialized tools (all kinds of databases, caches, etc). But Drools integrates with them by several different means, including pulling data from them on demand. I suggest you watch the presentation Drools and Large Data Sets by Alexandre Porcelli, at the bootcamp in Argentina back in June: http://blog.athico.com/2011/08/drools-jbpm5-sessions-argentina-june.html . The solution is obviously tailored to the problem in question (constantly dealing with 30 million rows of historical data with high performance if I remember correctly), as any solutions for any major problems are. In other words, using each tool for what they do best and build the puzzle from there. Regarding your second point, my understanding is that it has nothing to do with Drools per se. Does not matter if you implement your rules in Drools, in Java, or any other technology/language, you will have the same mapping problem. Solutions for that include a translator for the rules as you mentioned, or a translator for the data (smooks is a good open source one), or even use an adapter fact model that allows rules to work under the user friendly model, but execute actions on the engineering model. Hope it helps, Edson On Fri, Nov 11, 2011 at 5:53 AM, kapokfly ivan.jiang...@foxmail.com wrote: Not sure if these are also your questions you might have with the rule engine adoption, but we have seen 2 showstopper issues for drools adoption in our organization, or it is just because we are lack of some necessary knowledge. Issue 1) : pattern match for large data There are cases in a rule we need find matched records against a large population of data and perform some actions. Example: From 2 millions employee directory, find those belongs to Depart1 and with salary increased by 10% last year. Inserting all these employee's data into WorkingMemory is HUGE and mission impossible, we can't simply load the entire db into the working memory, do we have any other approach to accomplish or this is just not a use case can be resolved by a rule engine? Issue 2): Map business object to real engineering implementation What we are trying to do is to build a rule UI which a business user without engineering background of our codebase can work with; the business object we show on the UI might be different with the real engineering objects already in system, for example, on the UI a user might be able to able to see something like person.address.addressLine1 but the actually implementation might not be an object traverse implementation. Example, the model could be Long Person.addressId ; then referring to his address we need another lookup and we want to hide such complexity to the end user. The approach we took is to design a translation layer to translate the user input to drools rule string, is this the right approach? If not, any suggestions? I know Drools is mainly designed for developers but just trying to see if there is any way we can work it around. Thanks and looking forward to your recommendations. Ivan - Ivan, your Panda, forever -- View this message in context: http://drools.46999.n3.nabble.com/Showstopper-issues-for-Drools-Adoption-in-our-organization-tp3499367p3499367.html Sent from the Drools: User forum mailing list archive at Nabble.com. ___ rules-users mailing list rules-users@lists.jboss.org https://lists.jboss.org/mailman/listinfo/rules-users -- Edson Tirelli JBoss Drools Core Development JBoss by Red Hat @ www.jboss.com ___ rules-users mailing list rules-users@lists.jboss.org https://lists.jboss.org/mailman/listinfo/rules-users
Re: [rules-users] Showstopper issues for Drools Adoption in our organization
Hi Edson, Thanks for the informative reply. Will take a look and see how we can proceed and share our thought with you. Best wishes... Ivan - Ivan, your Panda, forever -- View this message in context: http://drools.46999.n3.nabble.com/Showstopper-issues-for-Drools-Adoption-in-our-organization-tp3499367p3499901.html Sent from the Drools: User forum mailing list archive at Nabble.com. ___ rules-users mailing list rules-users@lists.jboss.org https://lists.jboss.org/mailman/listinfo/rules-users
