A Schema-First Formalism for Labeled Property Graph Databases
Chandan Sharma, Roopak Sinha · 2019
Graph databases provide better support for highly interconnected datasets than relational databases. However, labeled property graph databases, which have become increasingly popular, are schema-optional, making them prone to data corruption, especially when new users switch from relational databases to graph databases. In this work, we provide a schema-driven formalism for graph databases. This formalism enables schema-driven loading of graph databases from other sources, such as relational databases. Also, this formalism enables schema-driven data analytics that allows for a more structured analysis of data stored in graph databases. Such analytics are based on a boilerplate approach allowing users who are not experts in the use of graph database query languages to carry out analytics efficiently. We showcase the utility of the proposed formalism by considering a case study from Airbnb for illustrating schema-based loading procedures. The proposed schema-driven analytics process is illustrated using another case study from an industrial cyber-physical systems standard. Overall, the schema-driven formalism provides several useful features, such as preventing both data corruption and long-term degradation of graph database structures.