Self-Detection and Comprehensive Learning-Based BRO for Cloud Workflow Scheduling Under Budget Constraints

To guarantee the diversified user QoS requirements, workflow scheduling in the cloud data centers still face challenges. In this paper, a Self-detection and Comprehensive Learning-based Battle Royale Optimization algorithm (SCLBRO) is proposed for scheduling workflows to optimize the makespan under budget constraints. Firstly, a Comprehensive Learning Strategy-based re-spawn mechanism is incorporated into the original Battle Royale Optimization (BRO) algorithm to improve the global search ability. Second, a local optimum detection method is designed by counting and evaluating the similar soldiers to reduce the possibility of falling into local optima. Third, an elite enhancement strategy is adopted to increase the search diversity for better balancing between exploration and exploitation. Extensive experiments are conducted on four well-known scientific workflows with different scales, and the results demonstrate that SCLBRO outperforms its peers in the success rate, convergence and solution quality.