Thanks Roberto for the questions. These are quite good questions. For AGG Group in your case, if none of the aggs could answer your query, the query will be executed against the base cuboid, which includes all dimensions. That also explains why the performance will degrade much, since the query will base on the post-processing at the running time.
TopN measure is a good way to process the HCD case. http://kylin.apache.org/blog/2016/03/19/approximate-topn-measure/ But if the filter condition has no order rule, the TopN could not be used. The AGG group will not help either. 2017-04-24 19:14 GMT+08:00 Roberto Tardío Olmos <[email protected] >: > Hi Kylin Community, > > I have some doubts about how aggregation groups works and HCD dimensions > good practices: > > 1. I have created a cube with two aggregation groups. In the first Agg > Group I have included Time and Company dimensions. In the second AGG Group > I have included Customer Dimension, with a cardinality about 1 million of > rows. This dimension is used with less frequency than the dimension at the > first agg group and filtering (some IDs) is always applied to it. > > After cube build, I can execute queries that combine Dimensions of the two > aggregations groups if a need. However, the query latency is quite poor > than when I define the three dimensions together at the same Agg Group. I > guess that is due to aggregation occurs during execution, because are no > precalculated like when the three dimensions are in the same AGG group. > > How the two aggregations groups are combined at query execution? I suppose > that the FK reference of any fact involved in the query result is stored > and known by two AGG groups. I would like to know more detail about how > this works. > > 2. If a have an HCD dimension that is little used in queries and always > applied Customer ID filtering to get data only for some customers. ¿Is is a > good practice to define it in a separated AGG group? > > Regards, > -- > *Roberto Tardío Olmos* > *Senior Big Data & Business Intelligence Consultant* > > Avenida de Brasil, 17, Planta 16. 28020 Madrid Fijo: 91.788.34.10 >
