A Hub-based Graph Management for Efficient Repetition Path Traversing

Abstract

Graph database (GDB) is one of the best approaches for enabling efficient management of relationships in the graph. It retains the shape of the graph to retrieve the required nodes through their interconnections. GDBs have an attractive characteristic called index-free adjacency, enabling users to traverse one edge at a fixed computing cost for vast and complex graph data. However, it is quite likely that GDBs cannot avoid reaching already-scanned nodes from different start nodes by repeatedly traversing edges that have a specific relationship type. This is because GDBs cannot consider the node degrees of each node in a graph. Consequently, when a GDB reaches a hub, the number of nodes that need to be arrived at by traversing different edges increases depending on the hub's adjacent nodes. Therefore, an inefficient process increases the cost of extracting a sub-graph in conventional GDBs. Accordingly, in this paper, we propose an efficient approach to repeatedly traversing edges belonging to a specific relationship type by distinguishing between hub s and other nodes. In particular, our approach manages repetition path s from each hub in addition to a list of the adjacent node in a conventional GDB; this GDB aims to enable us to manage a real-world network effectively.