Cloud computing faces challenges in task scheduling, which is crucial for cost-efficient execution and resource utili-zation. Current methods face computational complexity, especially in large-scale data centres. This paper proposes a novel approach that considers job dependencies and task execution times to reduce make-span, minimize energy con-sumption, andbalance resource loads. VMs are allocated based on workflow task requirements, using thresholds for task levels and durations to manage execution priorities. Tasks with higher dependencies and longer execution times are prioritized, ensuring efficient resource utilization and energy savings. The method employs queues for different task intensities, streamlining VM allocation by organizing tasks with additional metadata like intensities, arrival times, and deadlines. Historical scheduling logs (HSLs) are used to generate appropriate VMs, with new VMs created if no matching records exist in the HSLs. The proposed solution optimizes scheduling using an enhanced Multi-Tactic Harris Hawks Optimization (MTHHO) algorithm, which addresses the limitations of traditional HHO by incorporating Sobol sequences, elite opposition-based learning, and improved energy updating techniques to enhance population diversity, adaptability, and convergence accuracy while avoiding local optima using the Gaussian walk learning. The result shows that the proposed method of QoS performances attained less Makespan, energy consumption of 0.20, throughput of 2.4, and execution time of 16.75 with effectively allocated resources of 98% when compared to the previous methods in cloud computing. Therefore, the proposed heuristic-based MTHHO method balanced the load and allocated the resources effectively to improve QoS performances.