A cost-efficient QoS-aware model for cloud IaaS resource allocation in large datacenters

Abstract

Cloud Service Providers owning large datacen-ters are still challenged by how to serve large number of Infrastructure-as-a-Service (IaaS) requests and how to manage their massive repositories efficiently. Heuristic solutions are popular for the reasonable computational time required to calculate an optimal/near-optimal solution. Mathematical modeling based on Integer Linear Programming (ILP) technique is used to solve optimally the resource allocation problem. However, ILP technique is well-known that suffers from scalability issue, which makes it impractical for large datacenters. To overcome the scalability issue and provide an efficient solution in reasonable time, we propose a cost-efficient model acquainted with QoS requirements which makes use of large-scale optimization tools and introduces a Column Generation formulation for IaaS resource allocation in large datacenters (RA-IaaS-CG). Simulation results shows the superiority of RA-IaaS-CG over related heuristics and mathematical solutions. The proposed solution improves large datacenter resource utilization and outperforms other solutions in terms of scalability, resource utilization, and acceptance ratio.