We prove analytically that the performance of the new strategies is significantly better than certain limits suggested by existing literature. Under the additional (realistic) assumption that the load of each individual task is "small", we show that the processors can be almost fully utilized.

KEY WORDS: multiprocessor systems, hard real-time constraints, rate-monotonic scheduling, task assignment schemes, worst-case analysis