Hybrid Real-Time Task Scheduling Algorithm in Overload Situation for Multiprocessor System

Abstract

Real-time systems are reactive systems which should meet major constraints in scheduling tasks like time limitation and resources allocation for scheduling the task effectively when the system in overloaded condition. Failure of system in scheduling tasks when system is overloaded can result in catastrophic impacts. The goal of this research is to propose a task scheduling algorithm that able to perform better than traditional Earliest Deadline First (EDF) and minimize the overall completion time when the system in overloaded condition. The proposed scheduling algorithm is built based on three new improved scheduling algorithms namely: (1) Hybrid Particle Swarm Optimization (PSO) and Hybrid Invasive Weed Optimization (HPIO), (2) Enhanced Initial Swarm (EIS), and (3) Hybrid EDF, EIS and HPIO Optimization (HEDFPIO). The author proves that more successful tasks is scheduled by using HPIO in multiprocessor system in over loaded situation among PSO and ACO. The author uses EIS algorithm in order to improve local search in HPIO and have fair load balance among processors. Finally the author presents a new hybrid algorithm that combines HPIO, EIS and EDF which is called HEDFPIO, It is observed that we could achieve higher successful ratio in task scheduling and with shorter calculation time in overloaded situation.