Marika Erin McGhee
Mitigation of Performance Overhead Against Failed Self-Management
Within self-healing systems, the self-governing unit must know how to detect failures and respond appropriately to them. Each self-healing system may be comprised of two layers: a service layer and a self-management layer. The service layer provides application services, whereas the self-management layer is responsible for maintaining the success of the service layer. The self-management layer does not have any way of maintaining itself. In essence, it is assumed that the self-management layer does not fail in order for the entire system to continue. But it is very unlikely that the self-managing layer will never fail. This project addresses an approach to mitigating the performance overhead to the service layer whenever the self-management layer meets failures. The service layer monitors the status of self-management layer in self-healing systems, while the self-management layer checks the status of the service layer. When failures in the self-management layer are detected, the self-management layer is detached from the service layer in order to reduce the performance overhead. This project also analyzes the performance overhead of the self-management layer. A container-based component execution engine for robot systems is used for a case study.