Implementation of Fault-Tolerance Techniques for Real-Time Multiprocessor Scheduling

In this report I present some fault-tolerance techniques that has been implemented in a tool for real-time scheduling. The techniques implemented are based on some of the most established existing techniques found in the realtime community. The purpose of this work has only been to implement the techniques; later on they are supposed to be used for studies on fault-tolerance properties on real-time systems. The work basically consists of two parts: replication and reliability. The possibility to replicate tasks has been integrated into the scheduling tool. A few simple algorithms are implemented and more can easily be added. Support for tasks with different period times has also been integrated. A reliability measure, probability of no dynamic failure has been implemented. The measure is presented in conjunction with a brief introduction to basic fault tolerance theory. This quality measure could also be used to aid the scheduler for the purpose of improving reliability. Several reliab..