Dynamic Task Scheduling Algorithm based on Ant Colony Scheme

Many scientific applications running in Cloud Computing system are workflow applications that contains large number of tasks and in which tasks are connected by precedence relations. Efficient scheduling the workflow tasks become a challenging issue in Cloud Computing environments because the scheduling decides performance of the applications. Unfortunately, finding the optimal scheduling is known as NP-hard. Ant Colony Optimization algorithm can be applied to design efficient scheduling algorithms. Previous scheduling algorithms that use Ant Colony mechanism lack rapid adaptivity. This paper proposes a task scheduling algorithm that uses a modified Ant Colony Optimization. The modified version uses probability in order for ants to decide target machine. The proposed task scheduling algorithm is implemented in WorkflowSim in order to measure performance. The experimental results show that the proposed scheduling algorithm reduce average makespan to about 6.4% compared to a scheduling algorithm that uses basic Ant Colony Optimization scheme. Keyword-task scheduling, dynamic load balancing, Ant Colony Optimization, WorkflowSim, Pegasus workflows