The early assessment of system performance in distributed real-time systems

• Main Author: Giddings, Michael Anthony
• Published: Open University 2016
• Subjects: 004
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.698074

Distributed real-time process control systems are notoriously difficult to develop. They frequently overrun time schedules and break cost constraints. The problems are compounded where there are multiple development teams and stakeholders. Conventional model-driven development has been examined to see if it can be extended to resolve some of these problems. It may be possible to use early system design stages to identify performance issues which would otherwise not be identified until late in the development of the system. A functional model is proposed, in addition to those conventionally used for model-driven development, based on loosely coupled functional elements, to represent the behaviour of each system component. The model complements existing requirements and design specifications and addresses the combination of individual component abstractions to produce a complete system specification. The functional model enables the accurate prediction of system performance prior to the detailed design of each component. The thesis examines how performance can be calculated and modelled. An animator tool and associated code generator are used to predict system and component performance in a distributed aircraft navigation system. The use of the animator to support the system design prior to the generation of the component contract specifications and interface control documents provides a means of assessing performance which is accessible to domain experts and system designers alike. The model also enables the effects of requirements changes and component design issues on the system design to be assessed in terms of the system design to provide system wide solutions. This performance assessment model and animator compliments the existing 'fix-it-later' approach, reducing the chances of performance failure detected late during the system development process when they are most expensive to fix.