Processing scientific mesh queries in graph databases

Abstract

In this work-in-progress paper, we model scientific meshes as a multi-graph in Neo4j graph database using the graph property model. We conduct experiments to measure the performance of the graph database solution in processing mesh queries and compare it with GrAL mesh library and PostgreSQL database on synthetic and real mesh datasets. The experiments show that the databases outperform the mesh library. However, each of the databases perform better on specific query type, i.e, the graph database shows the best performance on global path-intensive queries and the relational database on local and field queries. Based on the experiments, we propose a mediator architecture for processing mesh queries by using the three database systems.