Multi objective workflow scheduling based on hybrid meta heuristics approach in cloud environment

Abstract

The significance of workflows has been notified in distributed computing environments as they help in analyzing the data in an organized way. The Workflows portray the working of scientific applications in various domains like biology, medicine, physics, and astronomy. This thesis investigates scheduling approaches for scientific workflow applications in IaaS cloud system. A comprehensive review of scheduling algorithms in cloud computing has been given to gain insights about recent development in this domain. Various heuristics and meta-heuristic techniques have been explored to identify the research gaps. A hybrid meta-heuristics-based method is proposed for scheduling dependent tasks that are capable of balancing the execution time, cost, and failure probability of the application. Initially, a Wind-driven optimization algorithm is implemented to generate the minimum schedule length while assigning the tasks to the resources. A wind-driven optimization algorithm is extended with a task priority phase to give a solution for the bi-objective scheduling problem considering. The extended work discusses the reliability model and generates a schedule with minimum makespan and maximum reliability by using a trade-off factor. A hybrid of Wind-driven optimization and Genetic algorithm has been implemented to obtain the optimized solution. In the presented approach, the schedule is generated with optimized cost under user-defined deadlines. The schedule obtained from GA is taken as the initial population of the WDO algorithm, which helps in improving the overall solution quality. The non-dominant sorting strategy is applied to achieve Pareto optimal solutions with makespan, cost, and reliability that enables the users to choose the best solution as per their preferences. The potential of the algorithm is illustrated using four different scientific workflows with varying computing requirements.