Optimization of Shortest-Path Search on RDBMS-Based Graphs

Kwangwon Seo,Jinhyun Ahn,Dong-Hyuk Im

doi:10.3390/ijgi8120550

Kwangwon Seo, Jinhyun Ahn + Show 1 more

Open Access

https://doi.org/10.3390/ijgi8120550

Copy DOI

Abstract

Calculation of the shortest path between two nodes in a graph is a popular operation used in graph queries in applications such as map information systems, social networking services, and biotechnology. Recent shortest-path search techniques based on graphs stored in relational databases are able to calculate the shortest path efficiently, even in large data using frontier-expand-merge operations. However, previous approaches used a sequential bidirectional search method that causes a bottleneck, thus degrading performance. The repeated use of an aggregate SQL function also degrades performance. This paper proposes a parallel bi-directional search method using multithreading. In addition, an efficient optimization method is proposed that uses B-tree indexing instead of an aggregate SQL function. Various experiments using synthetic and real data reveal that the proposed optimization technique performs more efficiently than conventional methods. As the size of data in practical applications continues to grow, these optimizations will enable the shortest path in a graph to be found quickly and accurately.

Highlights

Graph models are used in various fields for data that can be expressed using edges, which represent relationships between nodes
Algorithms to find the shortest path in a graph are an area of active research
This paper proposed optimization methods based on relational database management system (RDBMS) for finding a single source shortest path

Summary

Introduction

Graph models are used in various fields for data that can be expressed using edges, which represent relationships between nodes (which indicate a single point). Graph databases and the accompanying in-memory solution are unsuitable for large graphs in map information systems. They require ample memory to retain all nodes and edges in the graph. Conventional graph analysis requires transferring data out of the database into places where an analytic service (shortest-path search) is performed. For these cases, the user must know the database and the analytic framework. The advantages of an RDBMS lie in its scalability, stability, and ease of programming for graph management [2]

Methods

Results

Conclusion