Data Models

Ways to structure data

A data model is a set of concepts for describing data, relationships and constraints. The relational model, proposed by E. F. Codd in 1970, organises data into relations (tables) of tuples (rows) with attributes (columns) and is queried with SQL; it dominates business systems because of its strong theory, flexible queries and transactions. Older hierarchical and network models (IMS, CODASYL) are mostly historical. NoSQL models emerged for scale and flexibility: document stores (MongoDB) keep nested JSON-like documents; key-value stores (Redis, DynamoDB) map keys to opaque values for very fast lookups; wide-column stores (Cassandra, HBase) organise data by row key and column families for massive write volumes; graph databases (Neo4j) store nodes and edges for highly connected data such as social networks and recommendations. Many modern relational databases also support JSON columns, blurring the lines. The entity-relationship (ER) model is a conceptual design model used before choosing a logical model.

One customer in four data models

The same information shaped for different access patterns.

relational:   customers(id, name)          orders(id, customer_id, total)
              (42, 'Asha')                  (1001, 42, 1499)

document:     { "_id": 42, "name": "Asha",
                "orders": [ { "id": 1001, "total": 1499 } ] }

key-value:    "customer:42" -> '{"name":"Asha"}'

graph:        (Asha:Customer)-[:PLACED]->(o1001:Order {total: 1499})
              (Asha)-[:FRIEND_OF]->(Ravi:Customer)

Choose by access patterns

Relational is a safe default for most applications. Pick another model when a specific access pattern dominates: huge key lookups, deeply connected traversals, or flexible nested documents read as a unit.

Quick check: Who proposed the relational model of data, and when?

• Larry Ellison, 1979
• Peter Chen, 1990
• E. F. Codd, 1970
• Edgar Dijkstra, 1965