P&D Graph Cube: Model and Parallel Materialization for Multidimensional Heterogeneous Network
Xinyu Crystal Wu, Bin Wu, Bai Wang · 2017
We consider extending decision support facilities toward large-scale sophisticated networks, upon which multidimensional attributes are associated variety of entities, forming the so-called Multidimensional heterogeneous network. Multidimensional heterogeneous network has become important tool for modelling information networks, meanwhile, OLAP (Online Analytical Processing) has proven to beeffective tool on relational data, however, it is computationally an enormous challenge to manage and analyse Multidimensional heterogeneous network to support effective decision making and OLAP operations. In this paper, we enrich the semantics of the traditional OLAP and propose a P&D (Path and Dimension) Graph Cube Model framework, which can support multi-type queries and graph OLAP operations. Furthermore, on the basis of P&D Graph Cube Model, we divide the graph cube materialization into two parts described as Path related Materialization and Dimensionrelated Materialization. On Path Materialization, by taking account of structure summarization, we design the Pathrelated Materialization Algorithm based on the definition of relation path set, thus resulting in a more insightful and structure-enriched network. On Dimension Materialization, we provide a Graph Frag-shell Algorithm to compute graph shell fragments for fast high-dimension Graph OLAP. Finally, we implement the related algorithms on Spark. The results of experiments on real data set confirm the effectiveness and scalable of P&D Graph Cube Model with materialization algorithms.