Hi Kylin team,
We are trying to optimize Kylin's read performance. In our case, each query will only get data from a sub-range of all dates. So we tried to build multiple segments in the cube which is partitioned by dates, and each query should only read relevant segments. But according to our tests, no matter whether we partition the cube, the scan count is always the same for the same query. For example, assume we have 2 cubes. One is partitioned into 30 segments, and the other one only has one segment. The 2 cubes cover the same range of dates. If we run a query that contains date filter in it (which means for the 30 segments cube, the query should only read some of the segments), the scan count is the same between the 2 cubes. Also the execution time of the query is almost the same. This confuses us. We know partitioning in Kylin helps reduce the rebuild time, which is to improve the write performance. But why does partitioning make no improvement to the read performance? And is there any way t o optimize read performance in Kylin? Thank you!
