Exhaustive methods for determining whether all jobs complete by their
deadlines in systems that use modern priority-driven scheduling strate
gies are often infeasible or unreliable because the execution and rele
ase times of each job may vary. This article presents several worst-ca
se bounds and efficient algorithms for determining how late the comple
tion times of jobs can be in static and dynamic multiprocessor or dist
ributed systems.