You're welcome ShaoFeng, Thanks for sharing your experiences, the application case of the hybrid model that you indicate is also very useful.
Regards, From: ShaoFeng Shi [mailto:[email protected]] Sent: miércoles, 25 de abril de 2018 2:34 To: user <[email protected]> Subject: Re: Hybrid Cubes Document Thank very much to Roberto for the comment and suggestions! The feedback is helpful for us to keep this feature. I had thought it has no external user. For Manoj's question "Then Cube2 will be sufficient to get all query. Why do we need to combine?": We had the customer case that they deleted the source data after building into Cube (as the source data size is huge). Later when they add new dimension/measure, it is impossible to go back to rebuild the history data into the new Cube. So they use the hybrid model to combine history cube and new cube, then at least for the common dimension/measure, they can get an overall view. 2018-04-24 21:19 GMT+08:00 <[email protected] <mailto:[email protected]> >: Hi Kumar, “Then Cube2 will be sufficient to get all query. Why do we need to combine?” If you combine Cube 1 and Cube 2 using a Hybrid Model, queries that contains common columns (dimensions or measures) will be routed to hybrid cube, thus joining data from Cube 1 and Cube 2. I.e. This can be useful to avoid stopping service when we implement a new version of the cube. It is a concept difficult to explain. I recommend you to do a test, for example with the learning kylin project ,adding a new measure to the fact table. Regards, Roberto From: Kumar, Manoj H [mailto:[email protected] <mailto:[email protected]> ] Sent: martes, 24 de abril de 2018 10:21 To: [email protected] <mailto:[email protected]> Subject: RE: Hybrid Cubes Document Thanks for the explanation. #1 – Understood clearly whenever there is a change in structure. #2 – Still not clear how can Hybrid model be used as we need to defined 10 dimensions to Cube_2(5 Dimensions from Cube1). Then Cube2 will be sufficient to get all query. Why do we need to combine? Regards, Manoj From: [email protected] <mailto:[email protected]> [[email protected] <mailto:[email protected]> ] Sent: Tuesday, April 24, 2018 1:07 PM To: [email protected] <mailto:[email protected]> Subject: RE: Hybrid Cubes Document Hello Kumar, I try to answer your questions: 1. It can be useful in scenarios where i) we need to change cube definition optimizations (e.g. agg groups, mandatory dimensiones, rowkeys,…) or i) when we need to add columns or measures to a cube (the case I analyzed on document). Due to we cannot modify and existing cube without purging its data, create a new cube is the only option to perform changes. Due to cube re-building for historical data can be a hard process (time + resources), we can create a Hybrid Model over old and new cube and start to build new cube from last segment (e.g. date) of the old cube. After defining a Hybrid model queries that contains common columns (dimensions or measures) will be routed to hybrid cube, thus joining data from old and new cube. However, if we perform a query that includes one or more of new columns from new cube, it will be routed only to new cube. In mi opinion, this is a disadvantage of the hybrid model, because if we want to perform queries that include new dimensions, we have to rebuild all the historical for the new cube. 2. In your case Cube 1 will be old cube (historical data) and Cube 2 new cube (new data and columns). As there are not common columns between them, queries will be routed only to one of the cubes and not to hybrid cube. To apply hybrid model you must define cube 2 with 10 dimension ( 5 from cube 1 and the new ones from Cube 2). a. I do not know if it's your case, but If you are doing this for the purpose of dividing the construction of the cubes (e.g. building time optimization), you better use the Agg Group concept. Regards, Roberto From: Kumar, Manoj H [mailto:[email protected] <mailto:[email protected]> ] Sent: lunes, 23 de abril de 2018 9:39 To: [email protected] <mailto:[email protected]> Subject: RE: Hybrid Cubes Document Thanks.. Whenever we need to change the Cube Dimensions – To add new Dimension/Column, This Hybrid Model would be efficient??. As this doesn’t need to re-build the previous cube & save the building time. Also, Another use case, Can we use Hybrid Cube Model for below purpose - Cube 1- 5 Dimension - Cube 2 – Another 5 Dimension Hybrid Cube – Cube 1 + Cube 2 – Become 10 dimension Hybrid Cube is exposed to BI tool – User can see 10 Dimensions from both Cube combination?? Is this right statement? Regards, Manoj From: Billy Liu [[email protected] <mailto:[email protected]> ] Sent: Monday, April 23, 2018 8:48 AM To: user <[email protected] <mailto:[email protected]> > Subject: Re: Hybrid Cubes Document Hello Roberto, Thanks for this sharing. Would you like to publish it on the Kylin website? With Warm regards Billy Liu 2018-04-20 1:18 GMT+08:00 <[email protected] <mailto:[email protected]> >: Hi, Last days I was doing some researching about the use of hybrid cubes (model). However, I just found this document http://kylin.apache.org/blog/2015/09/25/hybrid-model/ published on Sep 25, 2015. Due to this fact, I wrote a little guide that aim to explain its use , possible use cases and current limitations. I share a document through the following link: https://drive.google.com/open?id=1qbvB1iONBcFMFE__SuF0ayq_l1_0vwXN Please, do not hesitate to correct me if you see something wrong. I have found this feature very interesting to mitigate the issues related to the problem of re building the entire Cube if we need to modify its definition. However, the hybrid model only combines the data from two cubes if a query uses only the common columns of this two cubes. I have analyzed this drawback in the document. I appreciate the help of the Kylin community and team, I hope this document helps. Best Regards, Roberto Tardío Olmos Senior Big Data & Business Intelligence Consultant Avenida de Brasil, 17 <https://maps.google.com/?q=Avenida+de+Brasil,+17&entry=gmail&source=g> , Planta 16.28020 Madrid Fijo: 91.788.34.10 http://bigdata.stratebi.com/ http://www.stratebi.com This message is confidential and subject to terms at: http://www.jpmorgan.com/emaildisclaimer <http://www.jpmorgan.com/emaildisclaimer> including on confidentiality, legal privilege, viruses and monitoring of electronic messages. If you are not the intended recipient, please delete this message and notify the sender immediately. Any unauthorized use is strictly prohibited. This message is confidential and subject to terms at: http://www.jpmorgan.com/emaildisclaimer <http://www.jpmorgan.com/emaildisclaimer> including on confidentiality, legal privilege, viruses and monitoring of electronic messages. If you are not the intended recipient, please delete this message and notify the sender immediately. Any unauthorized use is strictly prohibited. -- Best regards, Shaofeng Shi 史少锋
