posted on 2011-08-25, 12:12authored byPeter John Vidler
Distributed embedded systems can be very challenging to test—not only is there increased complexity in the design of the software and operating systems, but we may also have to evaluate different scheduling and fault-tolerance approaches for suitability. The result is that system-level testing will generally take place in hardware at a much later time than would be possible with a single-processor system.
Carrying out high-level testing at an early stage of the software development process offers a number of advantages in the common scenarios of changing and poorly specified requirements; this thesis aims to address these concerns.
A tool-supported process is introduced, which is capable of converting a single-processor prototype embedded system into various multi-processor equivalents, allowing the creation and use of a testable system at an earlier stage of development.
The process is then expanded to incorporate the automatic generation of source-code supporting redundancy, essentially allowing a developer to focus on writing code and leave the exploration of multi-processor scheduling and fault-tolerance mechanisms to the process described in this thesis.
These approaches are illustrated by means of empirical studies.