There are two main approaches to efficiently implement a data cube:
1. Compute and materialize cuboids incrementally from bottom up. Only materialize cuboids that are likely to be queried frequently to balance space and time costs.
2. Compute cuboids on demand by aggregating base cuboid data using pre-computed aggregation indexes. This avoids materializing the entire cube but answers queries efficiently using the indexes.
The optimal solution usually materializes a subset of cuboids and computes remaining cuboids on demand using aggregation indexes. The specific materialized cuboids depend on query workload and access patterns.
November 5, 2012 32
Problem: How to Compute Cuboids
Incrementally?