Raphtory: Modelling, Maintenance and Analysis of Distributed Temporal Graphs.

Abstract

Temporal graphs capture the development of relationships within data throughout time. This model ts naturally within a streaming architecture, where new events can be inserted directly into the graph upon arrival from a data source and be compared to related entities or historical state. However, the majority of graph processing systems only consider traditional graph analysis on static data, whilst those which do expand past this often only support batched updating and delta analysis across graph snapshots. In this work we de ne a temporal property graph model and the semantics for updating it in both a distributed and non-distributed context. We have built Raphtory, a distributed temporal graph analytics platform which maintains the full graph history in memory, leveraging the de ned update semantics to insert streamed events directly into the model without batching or centralised ordering. In parallel with the ingestion, traditional and time-aware analytics may be performed on the most up-to-date version of the graph, as well as any point throughout its history. The depth of history viewed from the perspective of a time point may also be varied to explore both short and long term patterns within the data. Through this we extract novel insights over a variety of use cases, including phenomena never seen before in social networks. Finally, we demonstrate Raphtory's ability to scale both vertically and horizontally, handling consistent throughput in excess of 100,000 updates a second alongside the ingestion and maintenance of graphs built from billions of events.