Graph Compression Storage Based on Spatial Cluster Entity Optimization

Abstract

Graph storage technology is confronted with an enormous challenge as far as the compact and complex graph-structure data. This phenomenon is derived from social networks with spatially intensive data. Since a hot event can cause the generation of a network cluster, which consists of a massive duplicate associated entities in the social networks, the space utilization and processing speed of graph data is obstructed. Therefore, it is necessary to design a graph storage mechanism specifically for the above data. In this paper, we propose a G raph compression S torage engine based on spatial C luster entity O ptimization ( GSCO ), which improves the native graph storage model through the proposed the many-to-one mapping structure and a H eat E volution E limination algorithm ( H2E ). Firstly, we define the spatial cluster entity formally and confirm the compressed storage objects. Then, we introduce the many-to-one relationship to transfer the mapping structure between the node and property. It compresses the data to raise the space utilization of the graph database. Finally, we propose the H2E algorithm that allows the representative nodes to be anchored an extended period in memory according to the heat evolution acceleration. It increases the hit rate and throughput and reduces the I/O operation by deleting the redundancy of data. Extensive experiments results show that the proposed GSCO storage model is better than Neo4j for reading and writing data in spatial clustering entity. It significantly promotes the effectiveness of graph operation, including the data loading, the common queries, and the clustering test.