“…The increasing amount of data and frequency of updates makes it impractical to compute graph results from scratch [24]. Existing incremental graph systems such as GraphInc [24], KineoGraph [25], and many others [3], [4], [23], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37] handle the dynamic nature of graph data by creating a series of snapshots from the incoming input and applying a graph algorithm on the latest snapshot. Given an input stream, these systems must efficiently manage graph snapshots and avoid redundant computations.…”