Understanding Graph Databases: Concepts, Trends, and Neo4j Example

Professor Zheng Weimin, an academician of the Chinese Academy of Engineering and a professor at Tsinghua University, published an article in People’s Daily titled “Seizing the Opportunity of Independent Innovation in Graph Databases,” urging domestic researchers and engineers to pursue independent innovation in both theoretical research and engineering development of graph databases.

Graph databases store and query data using graph structures, where nodes and edges represent entities and their relationships. This model treats relationships as first‑class citizens, allowing data to be connected naturally without a predefined schema.

The query language of a graph database typically includes a Data Definition Language (DDL) for creating, modifying, and deleting schema objects, and a Data Manipulation Language (DML) for querying and updating data. Such languages enable efficient retrieval of knowledge stored in a knowledge graph.

Because graph databases emphasize the interconnection of data, they offer inherent scalability and do not require a fixed data model, making them well suited for dynamic, relationship‑heavy applications.

Current trends show that, since 2014, interest in graph databases has far outpaced other database types. DB‑ENGINES statistics (up to 2021) indicate a steady rise in popularity, with Neo4j consistently ranking at the top of the graph‑database heat‑map.

Neo4j, a native graph database written in Java and Scala and released in 2007, implements the Property Graph Model. It supports ACID transactions, offers both community and commercial editions, and provides a built‑in web UI for data visualization. Users interact with Neo4j via the Cypher query language.

The Property Graph Model consists of nodes, relationships, and properties. Nodes can have labels and attributes; relationships have types and attributes, forming a directed, attributed graph.

As an illustrative example, a simplified enterprise knowledge graph is presented, containing five nodes (enterprises and persons) and four relationships (shareholder and supplier), each with associated properties.

Beyond Neo4j, other notable graph databases include Microsoft Azure Cosmos DB, OrientDB, ArangoDB, Virtuoso, and JanusGraph. A comparative chart highlights differences in licensing, language support, and deployment models.

According to DB‑ENGINES rankings (as of August 2021), the top‑20 graph databases are all developed outside China, underscoring the need for domestic innovation as emphasized by Professor Zheng.