The Relational Model


Relational Structure

The relational model is an advanced model for database systems. The main tenet of the model is the absolute separation of the logical view and the physical view of data. The physical view in relational is implementation dependent and not further defined.

The logical view of the data in relational is set oriented. A relational set is an unordered group of items. Each item is sub-divided into fields with atomic values. In a given set, all items have the same structure - the same number of fields and the same atomic data type for corresponding fields. Only the field values are different from item to item in a set.

A relational set is often modelled as a table. The items of a set are the rows of the table. The fields in the items are the columns. The columns in a table can have names. The rows are unordered and unnamed. A database consists of one or more tables plus a catalog (also represented by tables) describing the database.

Relational Power

The relational model defines a set of mathematical operations and constraints that can be applied to tables in databases. These comprise a fundamental theory that is provably correct. Relational operations and constraints are used to define business rules (user defined constraints). A DBMS provides mechanisms to support relational operations and constraints for user defined databases and business rules.

Because of the rigid separation of physical and logical views and mathematical underpinnings, the relational model offers great power and capability. The optimization possibilities are enormous in contrast to the limited optimization available for 3rd Generation Languages (3GLs) due to their ad-hoc definition. A relational language like SQL is descriptive, specifying the results desired rather how to obtain the results (procedural).

Using mathematical theorems, requested relational operations can be divided into components that can be processed by independent tasks running on one or more CPUs, even distributed machines. No other database model can offer such powerful parallel processing, not the hierarchical database model (IMS), the network database model (IDMS), the inverted file model or the object oriented database model (Gemstone).

Relational can support distributed data as well as distributed processing. Tables may be distributed, including parts of tables. The rows of a table can be partitioned and stored at distributed locations. Employee data may be partitioned between branch offices but appear as a single table at the headquarters. The columns of a table can also be distributed. Employee salary columns may be stored at the headquarters.

Unfortunately, the power of the relational model is often sabotaged by weak implementations. The next section, Relational Implementations and Compliance with the Model, looks at current relational database systems.

Return to Contents Page: Relational Databases, Do Users Benefit?     Return to Issues Page


Copyright © 2003 FFE Software, Inc. All Rights Reserved WorldWide