DSGA: A Distributed Segment-Based Genetic Algorithm for Multi-Objective Outsourced Database Partitioning

The outsourced distributed database is frequently used to tackle the large amounts of data in various smart city scenarios. The data partition technique is a significant research topic in the outsourced distributed database because it can directly affect the database performance in data exchange and...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Information sciences Jg. 612; S. 864 - 886
Hauptverfasser: Ge, Yong-Feng, Zhan, Zhi-Hui, Cao, Jinli, Wang, Hua, Zhang, Yanchun, Lai, Kuei-Kuei, Zhang, Jun
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier Inc 01.10.2022
Schlagworte:
Database partitioning
Distributed genetic algorithm
Evolutionary computation
Multi-objective optimization
Smart cities
Multi-objective optimization
Database partitioning
Smart cities
Distributed genetic algorithm
Evolutionary computation
ISSN:0020-0255, 1872-6291
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The outsourced distributed database is frequently used to tackle the large amounts of data in various smart city scenarios. The data partition technique is a significant research topic in the outsourced distributed database because it can directly affect the database performance in data exchange and sharing. Multiple objectives, including communication cost, load balance, and data privacy, should be considered during data partition. Previous approaches remain narrow in dealing with one of these objectives. However, the emphasis on any single objective cannot improve the entire performance of the outsourced distributed database. In this paper, a distributed segment-based genetic algorithm (DSGA) is proposed, which can protect data privacy as well as achieve the trade-off between communication cost and load balance. For privacy protection, a digit-based anonymity strategy is proposed based on the characteristic of attributes, which can maintain information integrity and achieve fuzzy identification. After that, a three-layer distributed framework is proposed for the multi-objective optimization to enhance the search efficiency and to achieve the trade-off between communication cost and load balance. Specifically, two segment-based operators, i.e., segment-based recombination and segment-based mutation, are proposed to sufficiently exchange evolutionary information to accelerate the convergence speed and to maintain population diversity to cover the whole multi-objective space, respectively. The performance of the proposed DSGA is verified in terms of solution accuracy and convergence speed. The effect of proposed strategies and operators in DSGA is also confirmed.
ISSN:0020-0255
1872-6291
DOI:10.1016/j.ins.2022.09.003