Summary: Task Reweighting under Global Scheduling on Multiprocessors #
Aaron Block, James H. Anderson, and UmaMaheswari C. Devi
Department of Computer Science, University of North Carolina at Chapel Hill
We consider schemes for enacting task share changes---a process called reweighting---on realtime multiprocessor platforms.
Our particular focus is reweighting schemes that are deployed in environments in which tasks may frequently request significant
share changes. Prior work has shown that fair scheduling algorithms are capable of reweighting tasks with minimal allocation
error and that partitioningbased scheduling algorithms can reweight tasks with better averagecase performance, but greater
error. However, preemption and migration overheads can be high in fair schemes. In this paper, we consider the question of
whether nonfair, earliestdeadlinefirst (EDF) global scheduling techniques can improve the accuracy of reweighting relative to
partitioningbased schemes and provide improved averagecase performance relative to fairscheduled systems. Our conclusion
is that, for soft realtime systems, global EDF schemes provide a good mix of accuracy and averagecase performance.
# Work supported by NSF grants CCR 0204312, CNS 0309825, CNS 0408996, and CCF 0541056. The first author was also supported by an NSF fellowship.
Realtime systems that are adaptive in nature have received considerable recent attention [3, 11, 13, 5]. In addition, multiproces
sor platforms are of growing importance, due to both hardware trends such as the emergence of multicore technologies and the
prevalence of computationallyintensive applications for which singleprocessor designs are not sufficient. In prior work [3, 5],
Block and colleagues considered the use of both fair and partitioningbased algorithms to schedule highlyadaptive workloads