Fixed-Priority Preemptive Multiprocessor Scheduling: To Partition or not to Partition
Paper i proceeding, 2000
Traditional multiprocessor real-time scheduling partitions a task set and applies uniprocessor scheduling on each processor. For architectures where the penalty of migration is low, such as uniform-memory access shared-memory multiprocessors, the non-partitioned method becomes a viable alternative. By allowing a task to resume on another processor than the task was preempted on, some task sets can be scheduled where the partitioned method fails.
We address fixed-priority scheduling of periodically arriving tasks on $m$ equally powerful processors having a non-partitioned ready queue. We propose a new priority-assignment scheme for the non-partitioned method. Using an extensive simulation study, we show that the priority-assignment scheme has equivalent performance to the best existing partitioning algorithms, and outperforms existing fixed-priority assignment schemes for the non-partitioned method. We also propose a dispatcher for the non-partitioned method which reduces the number of preemptions to levels below the best partitioning schemes.
partitioning
shared-memory multiprocessors
bin-packing algorithms
preemptive scheduling
fixed-priority scheduling
dynamic binding
multiprocessors
non-partitioned method
global scheduling
Real-time scheduling