These data models are based on application and user levels of data. They are modeled considering the logical structure of the objects in the database. This data models defines the actual relationship between the data in the entities.
For example, employee and department entities are related to each other by means of department. This minute level of relationship is defined in the record based data models. We will not be able get the mapping at this level in the object based data models. There it simply defines two entities are related. But the actual relationship between any two entities can be observed in record based data models.
There are 3 types of record based data models defined so far- Hierarchical, Network and Relational data models. Most widely used record based data model is relational data model. Other two are not widely used. Let us understand how they are different from each other.
Hierarchical Data Models
Imagine we have to create a database for a company. What are the entities involved in it? Company, its department, its supplier, its employees, different projects of the company etc are the different entities we need to take care of. If we observe each of the entity they have parent –child relationship. We can design them like we do ancestral hierarchy. In our case, Company is the parent and rests of them are its children. Department has employees and project as its children and so on. This type of data modeling is called hierarchical data model.
In this data model, the entities are represented in a hierarchical fashion. Here we identify a parent entity, and its child entity. Again we drill down to identify next level of child entity and so on. This model can be imagined as folders inside a folder!
In our example above, it is diagrammatically represented as below:
It can also be imagined as root like structure. This model will have only one main root. It then branches into sub-roots, each of which will branch again. This type of relationship is best defined for 1:N type of relationships. E.g.; One company has multiple departments (1:N), one company has multiple suppliers (1:N),one department has multiple employees (1:N), each department has multiple projects(1:N) . If we have M:N relationships, then we have to duplicate the entities and show it in the diagram. For example, if a project in the company involves multiple departments, then our hierarchical representation changes as below:
It helps to address the issues of flat file data storage. In flat files, data will be scattered and there will not be any proper structuring of the data. This model groups the related data into tables and defines the relationship between the tables, which is not addressed in flat files.
- Redundancy: – When data is stored in a flat file, there might be repetition of same data multiple times and any changes required for the data will need to change in all the places in the flat file. Missing to update at any one place will cause incorrect data. This kind redundancy is solved by hierarchical model to some extent. Since records are grouped under related table, it solves the flat file redundancy issue. But look at the many to many relationship examples given above. In such case, we have to store same project information for more than one department. This is duplication of data and hence a redundancy. So, this model does not reduce the redundancy issue to a significant level.
- As we have seen above, it fails to handle many to many relationships efficiently. It results in redundancy and confusion. It can handle only parent-child kind of relationship.
- If we need to fetch any data in this model, we have to start from the root of the model and traverse through its child till we get the result. In order to perform the traversing, either we should know well in advance the layout of model or we should be very good programmer. Hence fetching through this model becomes bit difficult.
- Imagine company has got some new project details, but it did not assign it to any department yet. In this case, we cannot store project information in the PROJECT table, till company assigns it to some department. That means, in order to enter any child information, its parent information should be already known / entered.
Network Data Models
This is the enhanced version of hierarchical data model. It is designed to address the drawbacks of the hierarchical model. It helps to address M:N relationship. This data model is also represented as hierarchical, but this model will not have single parent concept. Any child in the tree can have multiple parents here.
Let us revisit our company example. A company has different projects and departments in the company own those projects. Even suppliers of the company give input for the project. Here Project has multiple parents and each department and supplier have multiple projects. This is represented as shown below. Basically, it forms a network like structure between the entities, hence the name.
- Accessing the records in the tables is easy since it addresses many to many relationships. Because of this kind of relationship, any records can be easily pulled by using any tables. For example, if we want to know the department of project X and if we know SUPPLIER table, we can pull this information. i.e.; SUPPLIER has the information about project X which includes the departments involved in the projects too. Hence makes the accessibility to any data easier, and even any complex data can be retrieved easily and quickly.
- Because of the same feature above, one can easily navigate among the tables and get any data.
- It is designed based on database standards – ANSI/SP ARC.
- If there is any requirement for the changes to the entities, it requires entire changes to the database. There is no independence between any objects. Hence any changes to the any of the object will need changes to the whole model. Hence difficult to manage.
- It would be little difficult to design the relationship between the entities, since all the entities are related in some way. It requires thorough practice and knowledge about the designing.
Relational Data Models
This model is designed to overcome the drawbacks of hierarchical and network models. It is designed completely different from those two models. Those models define how they are structured in the database physically and how they are inter-related. But in the relational model, we are least bothered about how they are structured. It purely based on how the records in each table are related. It purely isolates physical structure from the logical structure. Logical structure is defines records are grouped and distributed.
Let us try to understand it by an example. Let us consider department and employee from our previous examples above. In this model we look at employee with its data. When we say an employee what all comes into our mind? His employee id, name, address, age, salary, department that he is working etc. are attributes of employee. That means these details about the employee forms columns in employee table and value set of each employee for these attribute forms a row/record for an employee. Similarly, department has its id, name.
Now, in the employee table, we have column which uniquely identifies each employee – that is employee Id column. This column has unique value and we are able to differentiate each employee from each other by using this column. Such column is called as primary key of the table. Similarly department table has DEPT_ID as primary key. In the employee table, instead of storing whole information about his department, we have DEPT_ID from department table stored. i.e.; by using the data from the department table, we have established the relation between employee and department tables.
Observe the table structures above. They are very simple to understand. There is no redundant data as well. It addressed major drawback of earlier data models. This type of data model is called relational data model.
This model is based on the mathematical concepts of set theory. It considers the tables as a two dimensional table with rows and columns. It is least bothered about the physical storage of structure and data in the memory. It considers only the data and how it can be represented in the form of rows and columns, and the way it can establish the relation between other tables.
A relational data model revolves around 5 important rules.
- Order of rows / records in the table is not important. For example, displaying the records for Joseph is independent of displaying the records for Rose or Mathew in Employee table. It does not change the meaning or level of them. Each record in the table is independent of other. Similarly, order of columns in the table is not important. That means, the value in each column for a record is independent of other. For example, representing DEPT_ID at the end or at the beginning in the employee table does not have any affect.
- Each record in the table is unique. That is there is no duplicate record exists in the table. This is achieved by the use of primary key or unique constraint.
- Each column/attribute will have single value in a row. For example, in Department table, DEPT_NAME column cannot have ‘Accounting’ and ‘Quality’ together in a single cell. Both has to be in two different rows as shown above.
- All attributes should be from same domain. That means each column should have meaningful value. For example, Age column cannot have dates in it. It should contain only valid numbers to represent individual’s age. Similarly, name columns should have valid names, Date columns should have proper dates.
- Table names in the database should be unique. In the database, same schema cannot contain two or more tables with same name. But two tables with different names can have same column names. But same column name is not allowed in the same table.
Examine below table structure for Employee, Department and Project and see if it satifies relational data model rules.
- Structural independence:- Any changes to the database structure, does not the way we are accessing the data. For example, Age is added to Employee table. But it does not change the relationship between the other tables nor changes the existing data. Hence it provides the total independence from its structure.
- Simplicity:- This model is designed based on the logical data. It does not consider how data are stored physically in the memory. Hence when the designer designs the database, he concentrates on how he sees the data. This reduces the burden on the designer.
- Because of simplicity and data independence, this kind of data model is easy to maintain and access.
- This model supports structured query language – SQL. Hence it helps the user to retrieve and modify the data in the database. By the use of SQL, user can get any specific information from the database.
Compared to the advantages above, the disadvantages of this model can be ignored.
- High hardware cost:- In order to separate the physical data information from the logical data, more powerful system hardwares – memory is required. This makes the cost of database high.
- Sometimes, design will be designed till the minute level, which will lead to complexity in the database.
Below table provides the comparison among the three models
|Hierarchical Data Model||Network Data Models||Relational Data Models|
|Supports One-Many Relationship||Supports both one to many and Many to Many relationship||Supports both one to many and Many to Many relationship|
|Because of single parent-child relationship, difficult to navigate through the child||It establishes the relationship between most of the objects, hence easy to access compared to hierarchical model||It provides SQL, which makes the access to the data simpler and quicker.|
|Flexibility among the different object is restricted to the child.||Because of the mapping among the sub level tables, flexibility is more||Primary and foreign key constraint makes the flexibility much simpler than other models.|
|Based on the physical storage details||Based on the physical storage details||Based on the logical data view|