Hi Dinghua, Kylin will estimate the cube size, while as both Kylin and HBase will use compression when writing to disk; so Kylin will apply a ratio on its origin estimated size. The ratio is collected based on experience; For normal data type, it is about 0.25, means the compressed size is about 1/4 of its origin size in theory; but for some measures like HyperLogLog or Bitmap (since they occupy much more memory, we call them "mem-hungry"), they can be compressed further, so it's default value is 0.05;
If you observed Kylin's default estimation isn't accurate in your environment, you can customize these two parameters to get a better result. 2016-12-15 12:14 GMT+08:00 Ding Dinghua <[email protected]>: > Hi, when I am reading the logic when estimating cuboid size, I found > the size has been multiply with the configured ratio, I can't get the > logic behind this, Could anyone tell me? Thanks > > -- > Ding Dinghua > -- Best regards, Shaofeng Shi 史少锋
