Navigation

The navigator is responsible for the content-aware data federation capabilities.

The input to the navigator is a logical request tree describing the precise semantics of the requested data while its output is the initial physical execution plan. The navigator exploits knowledge of content to eliminate the predominant majority of traditional multi-database joins.

It also has built-in support for common business analytics such as:

• time series comparisons,
• shares,
• and dimension-specific aggregation rules.

Rewrite

Once the navigator generates the initial physical execution plan, the rewrite phase of the compiler is responsible for distributed relational query optimization and optimal native SQL generation.

This phase covers:

• (i) Multidatabase join plan generation
• (ii) Function shipping
• (iii) Functional compensation analysis
• and (iv) Optimized SQL generation.

The Oracle BI Server’s join engine is seamlessly invoked when necessary. Join plans are constructed to maximize collective function shipping down to the source databases.

Optimal function shipping reduces loads on the source database and the network. The most important query processing elements to function ship include:

• GROUP BY and aggregation;
• Filters
• and Multi-pass SQL operations.

Equivalence Processing/Preserving

Aggregate and filter rewrites through a fragmentation design may push aggregates and filters through the tree (past operators such as joins, union alls, etc.) down to the database, thus reducing database load and network traffic. Both WHERE and HAVING filters may also be pushed to the database, depending on the GROUP BY clause.