[Paper Review] An improvement on fragmentation in Distribution Database Design Based on Knowledge-Oriented Clustering Techniques

Van Nghia Luong, Ha Huy Cuong Nguyen+1 more|arXiv (Cornell University)|May 6, 2015

Data Management and Algorithms5 references3 citations

TL;DR

This paper proposes a knowledge-oriented clustering approach to optimize horizontal and vertical fragmentation in distributed database design. By leveraging distance thresholds and enhanced similarity measures derived from classical methods, the algorithm reduces execution time and data fragmentation while maintaining stable complexity and matching classical results on small datasets.

ABSTRACT

The problem of optimizing distributed database includes: fragmentation and positioning data. Several different approaches and algorithms have been proposed to solve this problem. In this paper, we propose an algorithm that builds the initial equivalence relation based on the distance threshold. This threshold is also based on knowledge- oriented clustering techniques for both of horizontal and vertical fragmentation. Similarity measures used in the algorithms are the measures developed from the classical measures. Experimental results carrying on the small data set match fragmented results based on the classical algorithm. Execution time and data fragmentation significantly reduced while the complexity of our algorithm in the general case is stable.

Motivation & Objective

• To address the challenge of efficient data fragmentation in distributed database design.
• To improve upon existing fragmentation techniques by integrating knowledge-oriented clustering.
• To reduce execution time and data fragmentation while maintaining accuracy.
• To ensure stable algorithmic complexity across general use cases.

Proposed method

• The algorithm constructs an initial equivalence relation using a distance threshold derived from knowledge-oriented clustering.
• It applies similarity measures based on classical metrics, enhanced for database fragmentation tasks.
• The approach supports both horizontal and vertical fragmentation through unified clustering logic.
• The distance threshold dynamically adapts to data distribution patterns to improve clustering quality.
• Clustering is guided by domain knowledge to enhance relevance and reduce redundancy.
• The method is evaluated on a small dataset to validate performance and accuracy against classical algorithms.

Experimental results

Research questions

• RQ1 How can knowledge-oriented clustering improve the quality of fragmentation in distributed databases?
• RQ2 What is the impact of using a distance threshold on fragmentation efficiency and execution time?
• RQ3 Can the proposed method reduce data fragmentation while maintaining accuracy compared to classical approaches?
• RQ4 How does the algorithm's complexity behave under general database workloads?
• RQ5 To what extent does the integration of domain knowledge enhance clustering precision in fragmentation?

Key findings

• The proposed algorithm significantly reduces execution time compared to classical fragmentation methods.
• Data fragmentation is notably reduced, improving storage and query efficiency.
• The algorithm maintains stable complexity across various database workloads.
• Experimental results on a small dataset closely match those produced by classical algorithms.
• The integration of knowledge-oriented clustering enhances clustering accuracy and relevance.