Enhancing Execution in Kubernetes for Optimized Resource Utilization

Efficient workflow scheduling in cloud environments remains a critical challenge due to the trade off between performance (minimizing execution time) and operational cost. In this work, we propose a hybrid scheduling algorithm, HGSA, that leverages Heterogeneous Earliest Finish Time (HEFT) to establish task priorities and a refined Gravitational Search Algorithm (GSA) to optimize container assignments. Our approach seeds the GSA population with the HEFT solution and combines a simulated-annealing-inspired local search to escape local optima. We extend the makespan-only objective by incorporating a model that accounts for both computation and inter-container communication. Through various experiments on benchmark scientific workflows executed on a Kubernetes cluster, we demonstrate that our hybrid HEFT+GSA scheduler reduces average makespan compared to HEFT, vanilla GSA, and Particle Swarm Optimization (PSO). These results highlight the effectiveness of combining task-level heuristics with population-based metaheuristics for sustainable, high-performance cloud scheduling.