A Review of Engines for Graph Storage and Mutations

Abstract

With the continuous generation of big data, the need to structure a large amount of information is increasingly becoming a vital factor in extracting useful insights from raw data. Some of the technologies that emerged for this purpose are Graph Processing Systems that offer support for network analysis. Data can be collected and stored in a graph structure with vertices to represent entities and edges to represent their relationships, in order to reveal the correlation between different components e.g. to determine a group of users more likely to follow a certain Twitter account. In order to achieve high performance in Graph Analytics, graph processing engines exploit hardware resources and design efficient data structures to store graphs. Moreover, to track the evolution of graphs, systems need to support fast structural mutations i.e. addition/removal of vertices or edges. This paper provides a characterization of engines based on their hardware infrastructure, their graph storage and their support for graph mutations.