ROBUST: A Hardware Solution to Real-Time Overload

It has been proven that no on-line scheduling algorithm can guarantee a processor utilization greater than 25% under conditions of overload. From this result, it follows that there is no satisfactory software solution (within the paradigm of deterministic computation) to the problem of constructing overload-tolerant real-time systems, and we are forced to consider alternative remedies. In this paper, we describe an attempt at a hardware solution. In particular, we quantify the advantages of using faster hardware in real-time systems. An obvious advantage is that more work can be done in a shorter amount of time and the system is therefore less likely to enter overload. More interestingly, we prove that the performance of systems using faster hardware improves dramatically over systems using slower hardware if overload does occur. We present here a new scheduler, ROBUST, which efficiently uses faster hardware to prevent performance degradation under overload conditions