Energy-Efficient Management for Dynamic Consolidation of Virtual Machines in Cloud Data Centers under QoS Constraints

Dynamic Virtual Machine (VM) consolidation effectively reduces energy consumption and improves the utilization of resources in data centers. Reallocating VMs from an overloaded Physical Machine (PM) maximizes the utilization and energy efficiency with providing the required Quality of Service (QoS). The goal of consolidation of VMs ensuring an efficient utilization can be achieved through the use of VMs migration across different PMs. We present an overview of energy efficient design of cloud virtualized data centers to guide future design and development efforts. New VM placement algorithms are proposed and compared with a number of known VM placement algorithms. The algorithms are evaluated through simulations with real-world workload traces from more than a thousand VMs running on PMs located in more than 500 places around the world and it is shown that the proposed algorithms outperform the known algorithms. Moreover, a PM overload detection algorithm and a combination of PM overload detection algorithm and VM quiescing are proposed to maximize the time until migration, while meeting QoS goals. A number of benchmark PM overload detection algorithms are implemented using different parameters to compare with the proposed PM overload detection algorithm. We evaluate this algorithm through simulations with real world workload traces and results show that the proposed algorithm outperforms the benchmark PM overload detection algorithms under QoS constraints.