Early database systems modeled data as a hierarchy of records

The most popular database for business data processing in the 1970s was IBM's Information Management System (IMS). It is still in use today, running on OS/390 on IBM mainframes.

IMS relied on a simple hierarchical model, a similar data model to the JSON model used now by document databases. Data is represented as a tree of records nested within records.

IMS worked well for one-to-many relationships, but representing many-to-many relationships was difficult. It also did not support joins and left representation of relationships to users.

With IMS, Developers had to choose between denormalizing and duplicating data or resolving references manually from one record to another. The data model forced application developers to think a lot about the database's internal representation.

Also, the last three decades of software development have proven that it is important to develop and manage linkages between data entities. The relationships became so important that they have a major influence on the decision of the data model.

Several data models have now been proposed to overcome the drawbacks of the hierarchical model. The two most prominent were the relational model (which led to SQL) and the network model (also known as the CODASYL model after the committee that standardized the model). While relational databases took over the world and have ruled the database space for 25-30 years, the network model gained a lot of popularity in the 1970s but eventually faded into obscurity.