Analysis and Design of Dynamic Behaviour for Embedded Systems Using Policy-Based Design

Policy based computing is a technology which aims to abstract away decision making from the source code of a program. The technology provides exibility/changeability and adaptivity in terms of software behaviour. This is allows behaviour to be adapted in runtime. This thesis contributes with an investigation and analysis of the technology with a focus on realtime multicore embedded systems. The study identied several challenges in terms of policy based systems, and resulted in a suggested design. This thesis also includes a prototype implementation which showcases the usage of the suggested design. The implementation involves a policy engine which allows a low overhead, soft realtime performance, and context aware decision making for a targeted software component. The implementation is delimited to Linux based systems. It was concluded that realtime performance can be applicable to policy based systems, and that the suggested model gives high portability, and low over head. This makes the policy based approach highly applicable to embedded systems