<p>As an Internet-based computing model, cloud computing realizes the elastic scaling and efficient utilization of resources by centralizing computing resources (such as servers, storage and networks) to form resource pools and provide services to users on-demand. Task scheduling directly affects the operational efficiency, load balancing and energy consumption of the entire system. In order to improve the execution efficiency of task scheduling in cloud computing system, a Cycloid spiral <i>X</i> mayfly algorithm (CXMA) based on improved dance damping ratio and dance mode was proposed. Firstly, the elementary function is used to improve the dance damping ratio, which effectively improves the convergence stability of the algorithm and better balances the ability of global exploration and local development, so that the algorithm can locate the optimal solution more accurately while maintaining high search diversity. On the basis of improving the dance damping ratio, the basic mathematical function is used to improve the dance mode of the mayfly, and the search efficiency and solution accuracy of MA are significantly improved by optimizing the search behavior of the individual mayfly so as to improve the robustness and adaptability of MA. Through simulation experiments, the total cost, time cost, load cost and price cost of the system under large-scale and small-scale tasks are tested. Comparing the proposed CXMA with other swarm intelligence optimization algorithms, the experimental results show that the proposed CXMA has significant advantages in searching for the optimal task scheduling strategy. In terms of total cost, CXMA is 6.7% lower than ACO, 0.7% lower than CDO, 3.7% lower than WOA, 4.0% lower than BOA, 2.6% lower than AOA, 1.6% lower than SOA and 3.0% lower than RSO.</p>