Abstract | ||
---|---|---|
It has long been known that the global earliest-deadlinefirst (GEDF) scheduler is soft real-time (SRT) optimal for sporadic task systems executing on identical multiprocessor platforms, regardless of whether task execution is preemptive or non-preemptive. This notion of optimality requires deadline tardiness to be provably bounded for any feasible task system. In recent years, there has been interest in extending these SRT optimality results to apply to uniform heterogeneous platforms, in which processors may have different speeds. However, it was recently shown that nonpreemptive GEDF is not SRT optimal on such platforms. The remaining case, preemptive GEDF, has turned out to be quite difficult to tackle and has remained open for a number of years. In this paper, this case is resolved by showing that preemptive GEDF is indeed SRT optimal on uniform platforms, provided a certain job migration policy is used. |
Year | DOI | Venue |
---|---|---|
2017 | 10.1109/RTSS.2017.00037 | 2017 IEEE Real-Time Systems Symposium (RTSS) |
Keywords | Field | DocType |
global-EDF,uniform-multiprocessors,soft-real-time,tardiness | Metrical task system,Tardiness,Computer science,Multiprocessing,Job migration,Bounded function,Distributed computing | Conference |
ISSN | ISBN | Citations |
1052-8725 | 978-1-5386-1416-7 | 1 |
PageRank | References | Authors |
0.36 | 10 | 2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Kecheng Yang | 1 | 9 | 2.22 |
James H. Anderson | 2 | 3492 | 291.90 |