A New Method of Dynamic Horizontal Fragmentation for Multimedia Databases Contemplating Content-Based Queries

Felipe Castro-Medina,Jair Cervantes,Isaac Machorro-Cano,Asdrúbal López-Chau,Lisbeth Rodríguez-Mazahua,Giner Alor-Hernández,Mario Leoncio Arrioja-Rodríguez

doi:10.3390/electronics11020288

Abstract

The proper storage and management of multimedia data is a topic of great interest to industry and academia. Database fragmentation plays a fundamental role as a mechanism to guarantee cost reduction and improve response time performance in distributed data management environments. Multimedia database access patterns are constantly changing; due to this, it is important that the partitioning schemes also adapt to these changes. Dynamic fragmentation techniques offer this advantage and represent a reduction of the tasks that an administrator must perform and the complete autonomy to determine when to carry out a new fragmentation based on a cost model. This work proposes a new method of dynamic horizontal fragmentation for multimedia databases, including a way to contemplate content-based queries in the creation of new fragments. The use of content-based queries is on the rise, as multimedia elements are often presented within databases, and for this reason new fragmentation strategies must include this aspect to provide better-performing schemas. The method included in this research is placed within a current web application called XAMANA. We performed some experiments to demonstrate the effectiveness of our approach.

Highlights

Data fragmentation arises as a strategy to provide a new and adequate distribution unit in terms of performance
This research uses horizontal fragmentation to divide a relation along its tuples, considering a cost model that uses predicates to obtain each fragment and assign it to the site where it produces the lowest cost under the current workload
The approach presented fulfills the hypothesis of improving performance by reducing the execution time of operations by assigning the fragments to the sites where they are most used

Summary

Introduction

Data fragmentation arises as a strategy to provide a new and adequate distribution unit in terms of performance. Chunks in a Distributed Database Management System (DDBMS) allow data to be stored close to the points of use, and each site manages only a portion of the database, reducing I/O requests. Relational tables can be partitioned either horizontally or vertically. The basis of horizontal fragmentation is the select operator where the selection predicates determine the fragmentation, while vertical fragmentation is performed utilizing the project operator [1]. This research uses horizontal fragmentation to divide a relation along its tuples, considering a cost model that uses predicates to obtain each fragment and assign it to the site where it produces the lowest cost under the current workload

Objectives

Methods

Results

Discussion

Conclusion