Why Graph Databases Outperform Relational DBs for Social Network Queries

1. Unsuitability of Relational Databases

Relational data models have dominated since the 1980s, with major RDBMS products such as Oracle and MySQL. However, they struggle with modeling defects, horizontal scaling on large data volumes, and the exponential growth of data from users, systems, and sensors.

Data volume grows exponentially, creating big‑data storage and processing challenges.

Internet‑driven trends like social networks and intelligent recommendation increase urgency for scalable solutions.

These pressures have led to the emergence of many new databases, collectively called NoSQL databases, which can complement or replace traditional RDBMS.

2. NoSQL Data Models

NoSQL (Not Only SQL) encompasses a broad range of persistence solutions that do not follow the relational model and often avoid fixed table schemas and SQL JOINs, offering horizontal scalability.

NoSQL databases can be classified into four categories:

Key‑Value stores

BigTable‑style implementations

Document stores

Graph databases

Among these, graph databases have attracted the most attention in the past decade, as shown by db‑engines.com’s trend analysis.

To illustrate the urgency of using graph data, a small case study is presented.

3. What Matters Most in the New Internet Era?

Traffic is the lifeblood of companies: startups need traffic to attract investment, and large internet firms monetize traffic. Companies collaborate with mobile operators to offer various “free‑traffic” SIM cards, aiming to capture users.

Growth‑hacking strategies leverage algorithms to recommend connections (e.g., colleagues from the same university or hometown), creating multi‑degree social graphs.

4. Implementing Second‑Degree Friend Recommendation and Comparison

A simple relational schema includes a user table and a user_friends table. To find first‑degree friends, query user_friends by user ID; for second‑degree friends, first retrieve first‑degree IDs, then query their friends. Extending to third, fourth, or fifth degrees quickly becomes complex and inefficient.

Because the social network can be modeled as a directed graph, searching for N‑degree connections corresponds to graph traversal algorithms such as depth‑limited search, breadth‑first search, or Dijkstra’s algorithm. Graph databases handle these traversals naturally.

A performance test compares relational databases with the graph database Neo4j for finding friends up to depth 5 in a network of 1 million users, each with ~50 friends. Results show:

Depth 2: both databases perform adequately; Neo4j is ~2/3 the time of the relational DB.

Depth 3: relational queries exceed acceptable latency (≈30 s), while Neo4j stays under 1 s.

Depth 4: relational DB suffers severe delays; Neo4j remains within acceptable limits.

Depth 5: relational DB fails to complete; Neo4j returns results in ~2 s.

These findings generalize to other domains such as music recommendation, data‑center management, bioinformatics, and financial transaction analysis.

5. Unveiling Graph Databases

Graph databases store and query data using a graph structure of nodes and relationships (edges), rather than tables. Nodes have properties (key‑value pairs); edges have names, directions, and can also carry properties. G = (V, E); V = vertex (node); E = edge The name “graph database” reflects the underlying storage model, where systems like Neo4j keep user‑defined nodes and relationships as a graph, enabling efficient traversal from a starting node to discover connections.

The Property Graph Model further enriches this by allowing nodes and relationships to have attributes and multiple labels.

6. Neo4j – The Representative Graph Database

Popular graph databases include Neo4j, ArangoDB, OrientDB, and others, but Neo4j is the most active and widely used.

Neo4j can be installed easily on Windows; after installation, set the environment variables and start the server. The built‑in Neo4j Browser is accessed at http://localhost:7474/ (default Bolt URL bolt://localhost:7687, default user/password neo4j/neo4j – password must be changed on first login).

In the browser, Cypher queries create nodes and relationships. Example commands create two nodes and two relationships, which can be visualized directly in the Graph view.

7. Conclusion

Graph databases excel at handling massive, complex, interconnected data, delivering performance orders of magnitude higher than traditional relational databases. They are ideal for social networks, real‑time recommendation, banking transaction loops, and many other scenarios. Leading companies such as LinkedIn, Walmart, Cisco, HP, and eBay already use Neo4j, and adoption is growing worldwide.