City Research Online

Abstract

This thesis is concerned with the design and implementation stages of the development lifecycle of a class of systems known as hard real-time systems. Many of the existing methodologies are appropriate for meeting the functional requirements of this class of systems. However, it is proposed that these methodologies are not entirely appropriate for meeting the non-functional requirement of deadlines for work within these real-time systems. After discussing the concept of real-time systems and their characteristic requirements, this thesis proposes the use of a general transaction model of execution for the implementation of the system.

Whereas traditional methodologies consider the system from the flow of data or control in the system, we consider the system from the viewpoint of the role of each shared data entity. A control dependency is implied between otherwise independent processes that make use of a shared data entity; our viewpoint is known as the data dependency viewpoint. This implied control dependency between independent processes, necessary to preserve the consistency of the entity in the face of concurrent access, is ignored during the design stages of other methodologies. In considering the role of each data entity, it is possible to generate other viewpoints, such as the dataflow through the processes, automatically. This however, is not considered in the work.

This thesis describes a staged methodology for taking the requirements specification for a system and generating a design and implementation for that system. The methodology is intended to be more than a set of vague guidelines for implementation; a more rigid approach to the design and implementation stages is sought. The methodology begins by decomposing the system into more manageable units of processing. These units are known as tasks with a very low degree of coupling and high degree of cohesion. Following the system decomposition, the data dependency viewpoint is constructed; a descriptive notation and CASE tool support this viewpoint. From this viewpoint, implementation issues such as generating control flow; task and data allocation and hard real-time scheduling concerns, are addressed. A complete runtime environment to support the transaction model is described. This environment is hierarchical and can be adapted to many distributed implementations.

Finally, the stages of the methodology are applied to a large example, a Ship Control System. Starting with a specification of the requirements, the methodology is applied to generate a design and implementation of the system.

Publication Type:	Thesis (Doctoral)
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science Q Science > QA Mathematics > QA76 Computer software
Departments:	School of Science & Technology > Computer Science

Export

Downloads