Modeling and Analysis of Performance and Energy Consumption in Cloud Data Centers

Recently, the deployment of cloud data centers (CDCs) and the adoption of cloud technologies have transformed the way we do computation, storage and networking. Typically in a CDC, virtual machines (VMs) are allocated to physical machines. Estimating correctly the number of needed VMs to meet a given workload and QoS parameters is important for cost and resource efficiency. In this paper, we develop a queuing model to aid in studying and analyzing performance in CDC. We model the CDC platforms with an open queuing system that can be used to estimate the expected quality of service (QoS) parameters such as the throughput, the drop rate, the CPU utilization and the response time. In addition, we present an energy consumption model to study and estimate the energy consumption in the CDC. We give numerical examples to show how the proposed model estimates the number of needed VMs to meet a given level of QoS parameters. The results obtained from our analysis as well as the simulation models show that the proposed model is able to correctly and effectively estimate the number of VM instances required to achieve QoS targets under different workload conditions.