Relational-OLAP Tree

21 Feb 2020 - by @黄野

It’s been a while that I didn’t update my personal website, also speak or write in English. A long spring festival for most Chinese people due to the outbreak of Novel Coronavirus Pneumonia. Though I returned to work a little bit earlier than most people, I didn’t have to work very late as everyone prefer to reduce time spent with other people. What surprised me was my country could let 1.4 billion citizens wear masks over just one night.

As I have more time to stay at home, like today, it’s a good idea to record some memory pieces just in case.

Apache Kylin: A MOLAP-based Engine for big data. A cube is formed by a set of cuboids, and each cuboid is a multi-dimensional data model which consists of dimensions and measures.
Cuboid Spanning Tree: the root node contains all the dimensions which are decremented along with the growth of a cuboid spanning tree. Some nodes are redundant in real OLAP analysis scenarios, which could be cut off to reduce space and increase performance.
Tree Visualisation
- Each node can be clicked and the corresponding multi-dimensional table would be presented.
- An initial layout could be automatically generated by some base algorithms, such as circle, concentric, grid, radial, dagre, force, and so on. However, it could be more intelligent if some parameters which contain a real scene logic could be introduced.
Dimension Optimization
- Mandatory Dimensions: if not included, the calculation could be skipped.
- Hierarchy Dimensions: a good example is country -> province -> city, if no match, it could be regarded as an invalid combination.
- Joint Dimensions: some dimensions often appear together, such as restaurant - location.