RoC Prediction for Bi-Objective Cost-QoS Optimization of Cloud VoIP Call Allocations

In this paper, we present cloud VoIP scheduling strategies to provide appropriate levels of quality of service to users, and cost to VoIP service providers. This bi-objective problem is reasonable and representative for real installations and applications. We conduct comprehensive simulation on real data of sixty four strategies with dynamic prediction of the load. We show that our prediction rule that consider the number of Virtual Machines (VMs) running in the system improves the efficiency of traditional rate of change algorithm. It provides suitable quality of service and lower cost. Variations of VM startup time delays permit to evaluate the prediction rules under different scenarios and assess the robustness of all strategies.