The Global Limited Preemptive Earliest Deadline First Feasibility of Sporadic Real-Time Tasks

A Note on the Suboptimality of Nonpreemptive Real-time Scheduling

“…For the lower limit on S it is well known that any feasible task set having c max = 1 can be scheduled non-preemptively without the need to speed up the processor, hence trivially ≥ 1. Next, we show that the bound in (10) is tighter than the previous one presented in [8] [9] [10]. We prove this by showing that the limits of this bound is less than the bound presented in [10].…”

“…We prove this by showing that the limits of this bound is less than the bound presented in [10]. This is done by evaluating the limits of (10) for the same three extreme cases considered in [8], [9] and [10]. Proceeding:…”

“…Note that the schedulability conditions captured in (1), (2) and (3) are simple extensions of standard, known results to explicitly model the speedup factor S [8][9][10][11][12][13]. Various methods are known to bound L based upon the parameters of the task set, [13] provides a good discussion.…”

“…Consequently, we derive an upper bound on the processor speedup factor required to guarantee the schedulability of a feasible task set for npEDF scheduling. A similar approach to that employed in [8][9] [10] is taken, however a tighter bound is achieved.…”

“…Recently, Thekkilakattil et al [8][9] quantified the suboptimality of npEDF scheduling compared to EDF; any preemptively schedulable task set is also schedulable by npEDF with a processor speed not more than (4 ⁄ ) times faster, where represents the largest execution requirement of the task set and is the shortest relative deadline. However, it was later shown that this bound does not hold in the general case; a corrected representation was subsequently presented in [10] and given by:…”