A relational database is acollection of tables of such data which can be linked with some common culumns, primary key and foreign key in database term. For instance:
| EmpId | FName | LName |
| 1 | John | Q |
| 2 | Mary | James |
| 3 | Henry | Dew |
| EmpId | Type | Desc |
| 1 | Primary | This is primary description |
| 1 | Secondary | This is secondary description |
| 1 | Tertiary | This is tertiary description |
| 3 | Primary | This is primary description |
Here, these two tables can be connected with the common field EmpId. Also notice that employee 2 does not has any entry in second table. This way information can be separated without loosing its integrity. There would have been duplication and few blanks if these two tables were combined. This is a form of a relational database.
In above, the first table could be called Employee and the second, Description. In database term it would be written as:
Employee (EmpId, FName, LName)
Description (EmpId, Type, Desc)
Employee and Description are called "entity" and those in parenthesis are called attributes.
As with every design, one needs to decide what needs to be put into the table as shown above. How detailed do you want to be will decide on how many tables and how to group them.
Once this has been decided, linking those tables with common field would produce ERD and the documentation something like (for first table above):
| Table | Field | Data Type | Length | Description | Constraints |
| Employee | EmpId | Integer | Unique number that identifies an employee | Primary Key, Identity | |
| FName | varchar 50 | first name of an employee | |||
| LName | varchar 50 | last name of an employee |
Would be data dictionary where one defines data: name, type, length etc.
