HiNode: an asymptotically space-optimal storage model for historical queries on graphs

Management of historical information and preference in massive datasets

“…Table 6.6 showcases the space required by each model for each of the datasets in Table 6.4. 10 In all cases the ST model (which more closely represents the original HiNode vertex-centric model of [73]) achieves better space utilization than its counterparts.…”

“…In HiNode, each change is stored O(1) times, resulting in an asymptotically optimal total space cost. As an example, for the graph in Figure 6.1 the diachronic node for b at the time of G 2 would contain an Interval Tree for all changes related to b as well as three B-Trees for b's label, color and edge to e. Furthermore, due to the local handling of history, HiNode performs well on local queries and the authors further demonstrate that HiNode-G * is competitive on global queries as well [73]. In MT we elect to store time instances of change on a vertex or edge attribute ("timestamp") as opposed to explicit intervals since the underlying intervals can be trivially inferred.…”

“…Chapter 5 introduced the first purely entity-centric, and more specifically, vertex-centric model for maintaining graph historical data, termed as HiNode [73]. Its strongest point is that it builds upon a theoretical storage model that is asymptotically space-optimal.…”