Modeling the Scheduling Algorithms of Periodic Tasks and Analyzing its Performance on Real-Time Multiprocessor Systems


Abstract in English

The research presents molding and analytical study of several scheduling algorithms types in real-time multiprocessor systems. The performance of three scheduling algorithms have been analyzed : Earliest Deadline First Scheduling (EDF) , Least Laxity First Scheduling (LLF), and Earliest Deadline First until Zero Laxity Scheduling (EDZL). This paper considers the scheduling of n periodic, independed, and preempted tasks with implicit deadlines on a platform of m homogenous multiprocessor. It has compared in terms of the load on the processor (processor's busyness) , the number of migrations, and the number of preemptions and the number of times in which these algorithms did not succeed in achieving the time limits for tasks where the latter is considered the most important criterion in real time scheduling. It also considers scheduling growing task sets of periodic tasks starting from 4 task set up to 64 task set, in order to study the effect of increasing the number of tasks and processors also on the performance of the scheduling algorithms. As a result of research, the strengths and weaknesses in the performance of these three algorithms have presented. It is proposed the best type of real-time system to apply each algorithm according to the strengths of its performance.

References used

M. Joseph, “Real-time Systems: Specification, Verification and Analysis” , Prentice Hall,1996
J. A. Stankovic and K. Ramamritham, “Tutorial Hard Real-Time Systems”, IEEE Computer Society Press, 1998
M .Arezou and A. G.Selim, "Scheduling Algorithms for Real-Time Systems", School of Computing , Canada, 2005

Download