Business-Intelligence Queries with Order Dependencies in DB2

A research paper from the University of Toronto presents optimization techniques for business-intelligence queries using order dependencies (ODs). The work evaluates these techniques using an implementation in IBM DB2 V10 with the TPC-DS benchmark schema and IBM customer-inspired queries. Experimental results demonstrate significant performance gains, and the paper includes a polynomial-time algorithm for testing logical implication of ODs.

Use Cases

• Optimizing SQL queries with joins based on order dependency inference described in the paper.
• Simplifying query plans involving 'order by' or 'group by' clauses using the presented OD techniques.
• Improving performance of business-intelligence workloads by exploiting monotonicity properties in SQL functions.
• Testing logical implication for sets of order dependencies using the polynomial-time algorithm devised in the research.

Strengths

• Techniques are evaluated using the industry-standard TPC-DS benchmark schema.
• The paper presents a sound and complete inference algorithm for order dependencies over natural domains.
• Experimental results demonstrate a significant performance gain for query evaluation.

Limitations

• The actual dataset (queries, results) is not provided; only the methodology and results are described.
• Row count and specific data formats are unknown, limiting suitability assessment for direct data use.
• Column-level documentation is absent; field semantics must be inferred from the paper's description.

Provenance

Source

University of Toronto research paper by Jaroslaw Szlichta.

Collection Method

Research implementation and evaluation using IBM DB2 V10 and the TPC-DS benchmark.