A user-driven MILP framework for cost-efficient and performance-Aware Iaas resource allocation

Abstract

Cloud computing has indelibly reshaped contemporary IT infrastructure by offering scalable and economically viable resource provisioning. Infrastructure-as-a-Service (IaaS), a key component, provides flexible computing resources, yet optimizing their allocation balancing energy, latency, and provisioning costs remains a complex challenge. This research introduces a user-driven Infrastructure-as-a-Service (IaaS) optimization framework, leveraging Mixed Integer Linear Programming (MILP). This framework is meticulously designed for cost-efficient resource management and performance-aware virtual machine (VM) placement. A core feature is its facilitation of dynamic user-configurable parameters, specifically cost-prioritization weights (α, β, γ), endowing it with significant adaptability to diverse operational requisites. Through comprehensive simulation studies involving systematic variation of these weights and workload scaling, the framework’s efficacy is demonstrated in optimizing VM placement across distributed servers. This approach achieves substantial improvements in resource utilization and cost management while rigorously adhering to performance constraints. Ten distinct comparative analyses visually articulate the inherent trade-offs in this optimization landscape.