An Integrated Qos Dependency Aware and Multi Objective Metaheuristic Framework For Cloud Task Scheduling

Abstract

Cloud task scheduling involves heterogeneous virtual machines, differentiated service requirements, workflow precedence constraints, and competing time, energy, cost, and utilization objectives. We propose an integrated three-model framework. QTSF clusters independent tasks using normalized service and resource attributes, orders them by priority density, maps them to compatible virtual machines, and fixes load imbalance. DAWSM extends this framework to directed acyclic graphs through dependency-aware task clustering, a precedence-feasible discrete particle swarm representation, and constraint repair. SMMSF clusters virtual machines, normalizes four conflicting objectives, and adaptively refines schedules at scale. Formal propositions guarantee termination, unique assignment, precedence preservation, and bounded normalized fitness. QTSF achieves 5693 ms makespan, 92 percent load-balancing efficiency, and 88 percent utilization in the reported aggregate experiments; DAWSM achieves 95 percent dependency satisfaction, 4200 ms workflow time, 92 percent clustering efficiency, and 225 cost units; SMMSF achieves 4100 ms makespan, 210 kWh, 120 cost units, and 94 percent utilization. The three models incrementally address independent QoS scheduling, dependency-constrained workflows, and large-scale multi-objective scheduling.