Comparative Study of Bio-Inspired Algorithms for Task Scheduling

Abstract

Task scheduling in heterogeneous, resource-constrained computing environments—such as cloud, grid, and edge platforms—remains a challenging combinatorial optimization problem. Bio-inspired algorithms (BIAs) have emerged as strong candidates for near-optimal scheduling thanks to their robustness, adaptivity, and favorable compute-quality trade-offs. This manuscript presents a comparative study of five widely used BIAs—Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), Artificial Bee Colony (ABC), and Grey Wolf Optimizer (GWO)—applied to static, non-preemptive task scheduling under multi-objectivecriteria: makespan, energy, monetary cost, resource utilization, and SLA violations.

We formalize a unifying problem model, detail encoding and operators for each metaheuristic, and describe a controlled simulation design with heterogeneous virtual machines (VMs), synthetic workloads (independent tasks and DAGs), and realistic constraints (VM pricing tiers and power models). Statistical analysis across 30 independent runs per algorithm shows that GA and GWO consistently achieve the best makespan-cost-energy balance, with GA slightly superior on makespan and SLA, while GWO offers competitive performance with fewer parameters to tune. PSO performs strongly but is more sensitive to parameter settings; ABC offers a favorable convergence-stability profile; ACO shows strengths in balanced utilization but lags on makespan for large, highly skewed workloads. The findings highlight practical guidance on algorithm choice, parameterization, and stopping criteria for production-grade schedulers.