<p>Elastic scaling is a critical approach in cloud computing that improves resource utilization and reduces costs. However, existing scaling methods are predominantly based on overall real-time load fluctuations, with limited consideration given to the characteristics of cloud task execution time distributions, a key attribute of resource allocation. This paper introduces a fine-grained elastic scaling mechanism (FGESM) that accounts for the distribution of task duration, task priority, and execution progress. By leveraging a task-aware two-tier resource pool, our approach dynamically schedules and allocates cloud resources to maximize utilization efficiency while maintaining service quality. We also propose an execution-time-sensitive scheduling algorithm to optimize task-to-resource matching. Theoretical analysis provides performance guarantees, offering insight into the conditions under which our model operates optimally. Experimental results across various workloads demonstrate the effectiveness of our approach in improving task completion time, resource utilization, cost efficiency, and SLA compliance. FGESM shows improvements in cloud resource management in experimental evaluations, demonstrating potential to enhance service quality, reduce costs, and improve overall system performance.</p>