As of future work we intend to tighten our result by (i)
discarding less information during the computation of function
fi(t) and (ii) reducing the number of preemptions (i.e., the number
of iterations) considered in Algorithm 1 – it is indeed impossible
for a task to get preempted every Qi time units as assumed by
Algorithm 1 unless the periods of the other tasks enable such a
preemption scenario.