Usage of graph databases for social graph modeling

Abstract

This article is devoted to graph database management systems. The main characteristics and capabilities of those systems have been contemplated. The problems that may occur during the social network development have been selected to be solved using a graph data model. The most popular database management systems nowadays, namely, Neo4J, OrientDB and ArangoDB have been chosen for the study. Such characteristics of the selected databases as whether the software is proprietary or freely distributed, whether databases have up-to-date documentation or not, whether they are supported by developers, whether there is a community where you can get answers to your questions, and how much time is needed to master the database have been elaborated. The typical social network queries, when you need to receive results with a large depth of search quickly, have been developed using the query languages Cypher, OrientDB SQL and AQL used in Neo4J, OrientDB and ArangoDB respectively. The comparison of query execution speed has been performed for the selected databases. For this purpose, a graph that has 5000 nodes and 24900 connections has been built by implementing the Barabashi-Albert model for generating random-scale networks. The test tasks for finding friends of three users with the depth of 5 have been generated. The average time for each request has been estimated for several executions. The conclusions have been drawn and the recommendations regarding the selection of the best graph database for social network implementation have been made.