Random access with a distributed Bitmap Join Index for Star Joins.

Area of Science:

• Database Systems
• Distributed Computing
• Data Management

Background:

• Relational database performance heavily relies on indices, particularly for low-selectivity queries.
• Star joins are common, performance-intensive operations that benefit significantly from indexing at scale.
• Existing index-based solutions for star joins are nascent, creating a performance bottleneck.

Purpose of the Study:

• To introduce a novel distributed Bitmap Join Index (dBJI).
• To present a framework-agnostic strategy for efficient join predicate resolution.
• To address the performance challenges of low-selectivity queries in large-scale star join operations.

Main Methods:

• Development of a distributed Bitmap Join Index (dBJI).
• Implementation of a linear-time join predicate solving strategy.
• Empirical analysis using Hadoop technologies (HBase, Spark) and the Star Schema Benchmark (SSB).

Main Results:

• dBJI demonstrated significant performance improvements over full scan methods.
• Performance gains ranged from 59% to 88% for low-selectivity queries.
• The proposed solution effectively handles join predicates in linear time.

Conclusions:

• Distributed indices, like dBJI, can substantially boost low-selectivity query performance.
• The dBJI approach offers a scalable and efficient solution for large databases.
• Further development in distributed indexing is crucial for optimizing complex query operations.