This paper introduces a novel self-adaptive metaheuristic, the Fuzzy Adaptive Bayesian Scheduling Optimizer (FABSO), for optimizing task scheduling in cloud computing environments. FABSO extends the Interactive Search Algorithm (ISA) baseline by adding fuzzy inference and Bayesian learning modules, allowing adaptive balancing of exploitation and exploration during the searching process. The fuzzy module monitors agent interactions using nine-rule dynamic search-dependent decision matrices, whereas the Bayesian module adaptively adjusts the influence of memory through probabilistic inference to learn from experience. Bi-level control enhances FABSO’s responsiveness in handling dynamic cloud and heterogeneous workload scenarios. The performance of FABSO is evaluated based on several benchmark schedules using the CloudSim toolkit. Performance is measured against cutting-edge algorithms based on standard parameters, including task execution cost, makespan, and load variance. FABSO outperformed the baselines, achieving lower execution costs, faster convergence rates, and improved workload allocation across all evaluations. Additional statistical analysis also confirmed its superiority in accuracy and stability across a range of task sizes and VM configurations. The results demonstrate FABSO’s strength as an exemplary tool for adaptive runtime task scheduling in multi-objective cloud scenarios.
Wang et al. (Fri,) studied this question.