Particle swarm optimization algorithm for workflow scheduling based on spot instances

In response to the conflict between the completion time and execution cost of workflows with deadlines in cloud environments, as well as the instability of spot instances, an Adaptive Particle Swarm Optimization (APSO) algorithm is proposed. This algorithm utilizes the particle swarm optimization (PSO) algorithm to prioritize workflow tasks and dynamically allocate deadlines. During the resource allocation phase, spot instances are employed to reduce execution costs. In cases where spot instances are reclaimed, tasks are resubmitted using on-demand instances to mitigate the impact of spot instance reclamation on task interruption. Additionally, the algorithm introduces adaptive inertia weights and a Gaussian mutation mechanism based on PSO to balance the global and local search capabilities of the particle swarm, thereby avoiding premature convergence to local optima. Experimental comparisons with ICPCP, PSO, ProLis, and L-ACO algorithms show that, under the same deadlines, APSO can reduce workflow execution costs by up to 33.3%. The experimental results demonstrate that the proposed algorithm exhibits significant advantages in terms of workflow success rate and execution cost, effectively addressing the cost optimization problem constrained by deadlines in workflow scheduling.