Which are the different Classification of database management systems?

Answer 1

Database ClassificationDatabase systems are classified according to the way they represent data, i.e., on their provided data model. Broadly speaking, data models are classified intorecord-oriented and object-oriented models. There are three standardized record-oriented data models, namely, hierarchical, network,and relational. Although there is not a single object-oriented data model standard, as we discussed in 5.1.1 Object Relational Models, there are two objected-based approaches: namely, object-relational and object-oriented.

Let us summarize these models.

Hierarchical Data Model

The hierarchical model stores the data records in a tree structure with edges that connect the different records representing their relationships. Except for the root, all other nodes must have a parent node and each node may have either zero, or more child (dependent) records. Thus, in the hierarchical model, relationships between record types can be only one-to-one or one-to-many binary relationships.

IBM's IMS is possibly the only hierarchical DBMS in use today. There are only a few places where it is being used. In IMS, many-to-many relationships are handled by storing the data in two trees with opposite dependencies. Recursive relationships are handled in the same way. This means that the database contains redundant data, which is a waste of space, which makes the database structure inflexible to data change, and which makes the update operations more complex. Basic referential integrity of parent-child relationship is enforced by the DBMS.

The hierarchical model does not support declarative query processing. Retrievals are achieved through navigationembedded within a high-level programming language. The navigation is a tree pre-order traversal that starts from the root and proceeds down the tree, accessing the subtrees in order from left to right. This means that users know and operate directly on the physical schema. Hence, the hierarchical model supports a low level of abstraction compared to relational and object-oriented models, and it also offers only weak security and integrity.

Network Data Model

The network model is similar to the hierarchical one in that it eliminates data redundancy in a database by storing the data in a graph (or network) structure rather than in a tree structure. In this way, recursive and circular relationships, for example, can be expressed with edges connecting two records of the same type.

The basic one-to-many relationship is represented by connecting the dependent records together in an ordered linked list with the parent record, which is called the owner, at the head of the list. Records are permitted to contain more than one link. This capability is used in expressing many-to-many binary relationships between two record types. Many-to-many relationships are expressed with the help of intermediate records that can be directly linked with records of both the related record types, providing an indirect link between these records. (The use of intermediate records is called decomposition.) Basic referential integrity constraints are enforced by the DBMS.

Similar to the hierarchical data model, the network model does not provide for declarative query processing. Access to records is achieved by navigation through the network structure using a set of predefined commands. These one-record-at-a-time commands must be embedded in a general-purpose programming language. Unlike records in the relational model, records in the network model can have fields, which are either single value, called simpleattributes, or multi-valued, called vector attributes.

Relational Data Model

In the relational model, data is stored in tables that completely hide the physical organization of data. Thus, compared to the hierarchical and network models, the relational model supports a high level of abstraction and a flexible database structure. The database structure, i.e., the schema, can easily be changed to accommodate new data and new access requirements.

Relationships of any cardinality (1:1, 1:N, and N:M) and degree (binary, recursive, or higher order) can be expressed by means of foreign keys that are logical pointers. Binary 1:1 and 1:N, and binary recursive relationships, can be expressed directly, whereas high-order and N:M relationships can be expressed by decomposition, i.e., with the use of an intermediate table.

In addition to basic referential integrity, the relational model supports semantic integrity constraints by means of checks, assertions, and triggers.

Finally, as we have discussed in some detail, the relational model supports declarative, interactive query languages such as SQL and QBE. Furthermore, it can be embedded in almost any programming language, and therefore it supports the development of database application programs.

Object-Oriented Data Model

The object-oriented model supports a high level of abstraction and exhibits the same flexibility in specifying data relationships as the relational model. Relationships in the object-oriented model are expressed as references to object IDs. However, the object-oriented model offers more flexibility in specifying the schema of the database by means of complex types, user-defined types, user-defined functions, and object inheritance.

Unlike the object-relational model, object-oriented data models are strongly tied to an object-oriented programming language (such as C++ and Smalltalk) acting as their data manipulation language. (In this respect, object-oriented models resemble the hierarchical and network models.) The object-oriented language is extended to support persistent objects whose values are stored in the database and can be shared with other programs. By manipulating transient and persistent objects uniformly, object-oriented models eliminate the so-called impedance mismatch; that is, object-oriented models can thus eliminate any incompatibility between the declarative database manipulation language and the procedural application programming language that is due to difference in programming paradigms and type systems. Thus, object-oriented databases simplify the task of programming by supporting a single level storage and a computationally complete DDL/DML.

In order to support declarative and interactive query processing, the ODMG standardization treats the object-relational model as a special object-oriented model, implementing its declarative query language.

Answer 2

Classification of DBMSs

Data Model Classification

Relational data model

Object data model

Hierarchical data model

Network data model

Object relational data model

Number of Users

Single User systems

Multi User systems

Number of Sites

Centralized - data is stored at single site.

Distributes - database and DBMS software stored over many sites connected by network

Homogeneous - use same DBMS software at multiple sites.

Cost

Low-end systems under $3000

High-end systems, over $100,000