Deadline Constrained Modified Min-Min algorithm for Cloudlet Scheduling

Abstract

Cloud computing provides scalable and cost-effective computing services through virtualization and distributed resource sharing. Efficient task scheduling and load balancing are important challenges in cloud environments, as improper cloudlet allocation may increase Makespan, reduce resource utilization, and cause deadline violations. Most existing scheduling algorithms mainly focus on minimizing completion time and often ignore important factors such as current VM load, computing capability, expected execution time of cloudlets on different virtual machines, and task deadlines. This paper proposes a Deadline-Constrained Modified Min-Min (DCMM) scheduling algorithm for efficient cloudlet allocation and dynamic load balancing in cloud computing environments. The proposed algorithm initially allocates cloudlets using a modified Min-Min approach by selecting the VM with minimum completion time. A rescheduling mechanism is then applied to balance workload by migrating tasks from heavily loaded virtual machines to lightly loaded ones whenever better completion time can be achieved. The algorithm also considers task deadlines and reduces the number of missed deadlines during scheduling. The proposed method is implemented using the CloudSim simulation framework and compared with the Smarter Round Robin (SRR) algorithm. Experimental results show that the DCMM algorithm achieves better Makespan, improved load balancing, higher resource utilization, and more tasks meeting their deadlines.