Load sharing in fault-tolerant real-time systems

We propose a dynamic load sharing fault-tolerant algorithm for real-time systems based on the comparison model of computation. Each real-time task is assigned to a group of processors for execution and the results produced are compared to determine the correct output and to locate the failures. When a task arrives to a processor, the processor tries to form a group with its neighbors to guarantee its deadline. If the task cannot be guaranteed locally, it will be transferred to another processor that has maximum probability of forming a group with its neighbors. We give the details of the algorithm for torus interconnection networks.