Multi-core-Aware VM Placement Algorithms Using Constraint Programming in Cloud Environment

One of the key challenges in a cloud data enter is how do we place virtual machines on physical machines. Mapping virtual machines onto physical machines is known as virtual machine placement. The optimal virtual machine placement improves resource utilization and power consumption. The parameters such as resource wastage, overloaded physical machines and number of active physical machines are improved by properly mapping the virtual machines. The improper mapping between virtual machines and physical machines increases the overloaded and under-loaded physical machines. Due to this, the resource utilization is reduced and more number of active physical machines is introduced. Therefore, physical machine consolidation techniques are needed to rearrange the virtual machines from overloaded and under-loaded physical machines. To place a virtual machine on a physical machine, various resource constraints are checked to see whether the physical machine is having the resource capacity demanded by the virtual machine or not. If the physical machine is not having sufficient resource capacity, then the next physical machine is checked. In most of the existing research works on virtual machine placement algorithms, the resource demands of the virtual machines are compared against the total resource capacity of physical machines. It is a simplistic one-dimensional problem leads to suboptimal results like no dedicated cores and overloaded cores. Here, there is a possibility of assigning the virtual machines beyond the CPU capacity of the pCPU. Therefore, a lot of pCPUs will be overloaded, and hence, a lot of physical machines will be overloaded. Because of this, the virtual machines may not get enough resource to run the applications on it. So, the performance of the overall system is degraded. Also, the virtual machines are not assigned to a dedicated pCPU of a physical machine.