The concept of cloud computing has evolved to facilitate tasks by dynamically distributing virtual machines. Task scheduling is a major concern in cloud computing, presenting a challenge for service providers. To address this, the study utilizes a simulator called “CloudSim” to explore load balancing in cloud computing and potentially contribute to the development of load balancing and task scheduling communities. The suggested solution considered the makespan parameters to address the issue with current metaheuristic methods. However, scheduling in cloud settings is a challenging problem as it is almost NP-complete. Consequently, several approximation-based solutions, notably those inspired by swarm intelligence, have been created. To achieve optimal results, we propose a mutation-based glowworm swarm optimization method. Our primary contribution is a method for reducing makespan across various data centers for a single job set. We model the new technique and evaluate its efficacy with different numbers of cloudlets and data centers using the CloudSim toolkit package. The simulation results demonstrate that our proposed load balancing technique significantly reduces the makespan compared to existing heuristic-based models.