Summary: An EDFbased Scheduling Algorithm for Multiprocessor
Soft RealTime Systems #
James H. Anderson, Vasile Bud, and UmaMaheswari C. Devi
Department of Computer Science
The University of North Carolina at Chapel Hill
We consider the use of the earliestdeadlinefirst (EDF) scheduling algorithm in soft realtime multiproces
sor systems. In hard realtime systems, a significant disparity exists between EDFbased schemes and Pfair
scheduling (which is the only known way of optimally scheduling recurrent realtime tasks on multiprocessors):
on M processors, all known EDF variants have utilizationbased schedulability bounds of approximately M/2,
while Pfair algorithms can fully utilize all processors. This is unfortunate because EDFbased algorithms entail
lower scheduling and taskmigration overheads. In work on hard realtime systems, it has been shown that
this disparity in schedulability can be lessened by placing caps on pertask utilizations. In this paper, we show
that it can also be lessened by easing the requirement that all deadlines be met. Our main contribution is a
new EDFbased scheme that ensures bounded deadline tardiness. In this scheme, pertask utilizations must be
capped, but overall utilization need not be restricted. The required cap is quite liberal. Hence, our scheme
should enable a wide range of soft realtime applications to be scheduled with no constraints on total utilization.
We also propose techniques and heuristics that can be used to reduce tardiness. To the best of our knowledge,
this paper is the first to examine multiprocessor EDF scheduling in the context of soft realtime systems.
Keywords: Multiprocessor systems, soft realtime, earliestdeadlinefirst scheduling.