As we know:
The default maximum size of kylin dimension is 63, which is specified by the kylin.cube.rowkey.max-size value in the configuration file.
The default maximum cubic combination of kylin cuboids is 32768, which is specified by the kylin.cube.aggrgroup.max-combination value in the configuration file;
We can use many various methods to reduce cuboid's combination
In our system, we used more than 80 dimensions, the size combination is just 1000 combinations, is very fixed.
After setting kylin.cube.rowkey.max-size to 128, if I save a cube which has more than 64 dimensions, kylin throws exception: "Cuboid -1 out of range 0-9223372036854775806".
After analysis, the cuboid in kylin is identified by the N-bit binary number, and the cuboid is saved with long type. If a cube has more than 64 dimensions, the cuboid value exceeds the maximum value of the long type and the error is reported.
Can kylin support this scene which has a limited combination of category dimensions (less than 1000) but has large number of dimensions (more than 80)?
Can kylin save the cuboid value using BigInteger type? What risk if kylin save cuboid uses BigInteger type?
This message was sent by Atlassian Jira