Multidimensional Data Model

• Collection of numeric measures, which depend on a set of dimensions
• Dimension hierarchies
  • For each dimension, the set of values can be organized into a hierarchy

Implementation

• MOLAP
  • Multidimensional data is stored physically in a multidimensional array
  • Fast array operations (reading), expensive set operations
• ROLAP
  • Multidimensional data is stored as relations
  • Fast set operations (set, union, etc), meh array operations
  • Implementation
    • The main relation (which relates dimensions to a measure) is called the fact table
    • Each dimension an have additional attributes and an associated dimension table
    • Fact tables are much larger than dimensional tables
  • Star Schema
    • One fact table with many dimension tables
    • Because the fact table is in the middle and links all of the dimension tables together

Marginals

• Compresses over a dimension of a hypercube
• Aggregates over dimension(s) and stores as a slice
• Visualized as next to the cube
• Speeds up other queries

Cube Operator

• If there are k dimensions, there ark 2^k possible sql group by queries that can be generated through pivoting on a subset of dimensions
• GROUP BY attr1, attr2, ... WITH CUBE
• Equivalent to rolling up on all subsets of {attr1, attr2, ...}
• Generates a table with the schema of {attr1, attr2, ...} where the attributes that were aggregated over for each row is null