A cost, time, energy-aware workflow scheduling using adaptive PSO algorithm in a cloud–fog environment

Recent years have seen an exponential rise in data produced by Internet of Things (IoT) applications. Cloud servers were not designed for such extensive data, leading to challenges like increased makespan, cost, bandwidth, energy consumption, and network latency. To address these, the cloud–fog environment has emerged as an extension to cloud servers, offering services closer to IoT devices. Scheduling workflow applications to optimize multiple conflicting objectives in cloud fog is an NP-hard problem. Particle Swarm Optimization (PSO) is a good choice for multi-objective solutions due to its simplicity and rapid convergence. However, it has shortcomings like premature convergence and stagnation. To address these challenges, we formalize a theoretical background for scheduling workflow applications in the cloud–fog environment with multiple conflicting objectives. Subsequently, we propose an adaptive particle swarm optimization (APSO) algorithm with novel enhancements, including an S-shaped sigmoid function to dynamically decrease inertia weight and a linear updating mechanism for cognitive factors. Their integration in cloud–fog environments has not been previously explored. This novel application addresses unique challenges of workflow scheduling in cloud–fog systems, such as heterogeneous resource management, energy consumption, and increased cost. The effectiveness of APSO is evaluated using a real-world scientific workflow in a simulated cloud–fog environment and compared with four meta-heuristics. Our proposed workflow scheduling significantly reduces makespan and energy consumption without compromising overall cost compared to other meta-heuristics.