2.5. Summary

An information system for a given application may be viewed from at least four levels: conceptual, logical, physical, and external. At each level the formal model or knowledge base comprises a schema that describes the structure or design of the UoD and a database that is populated with the fact instances. Each schema determines what states and transitions are permitted for its database(s).

The conceptual schema does this in terms of simple, human-oriented concepts. The logical schema groups information into structures supported by the generic logical architecture (e.g., relational). The physical schema specifies the physical storage and efficient access structures (e.g., indexes) for the specific DBMS or platform being used to implement the application. For the same global conceptual schema, different external schemas can be constructed for different user groups depending on what information is to be accessible and how the information is to be displayed.

A conceptual schema comprises three main sections: fact types, constraints, and derivation rules. Asserted or base facts are primitive (i.e., they are not derived from other facts). Derived facts are computed or inferred by applying derivation rules to other facts. In declaring fact types, we indicate what kinds of object there are, how these are referenced, and what roles they play. Each role in a relationship is played by only one object type.

The simplest kind of object is a value (e.g., character string). Entities are real or abstract objects that are identified by their relationships to other objects (typically values). For example, a country may be identified by its name.

Static constraints determine the allowable populations of fact types. Dynamic constraints restrict transitions between these populations. For example, each city is the capital of at most one country, and no adult may become a child.

Each fact in a conceptual database is typically elementary. The addition or deletion of a fact is a simple update. In a compound transaction, several simple updates may be included; in this case, constraints apply only to the net effect of the complete update sequence, not to each individual update. Updates and queries on a conceptual database or conceptual schema are responded to by the conceptual information processor (CIP).

Each DBMS conforms to a logical data model (e.g., network, hierarchic, or relational). If a relational DBMS is chosen for the implementation, the conceptual schema is mapped to a relational schema, or relational database schema. Here all the stored facts are placed in named tables, with named columns and unnamed rows. Each row of a relational table typically corresponds to one or more elementary facts. Each cell (row-column position) contains at most one value. Within a table, no row may be duplicated. The population of a relational table may be changed by inserting, deleting, or modifying a row. Most conceptual constraints and derivation rules can be expressed either within the relational DBMS language or by interfacing to another language.

The information systems life cycle typically involves the following stages: feasibility study; requirements analysis; conceptual design of data and processes; logical design; basic physical design; basic external design; prototyping; implementation of production version; testing and validation; release of software, documentation, and training; and maintenance. Feedback from a later stage may indicate a need to rework earlier stages.

Careful attention to earlier stages reduces the overall effort involved, as does reuse of design strategies used previously on similar applications. For large projects, an iterative approach to development is generally recommended, where the life cycle is applied to one component before moving onto another component.

Various frameworks for information system architecture exist. The Zachman framework is composed of five rows that view the system from different perspectives (scope, enterprise model, system model, technology model, and details) and six columns with a different focus (data, function, network, people, time, and motivation).

The extended four-schema framework uses the same six columns, but has six rows for conceptual, external, logical-persistent, logical-transient, physical-persistent, and physical-transient perspectives. Other well-known system architectures include TO-GAF and MDA.