Uma abordagem evolucionária para o posicionamento dinâmico de máquinas virtuais em datacenters privados, aplicando sobrecarga de recursos.

Abstract

The advent of cloud computing technologies has helped companies with solutions to maintain their computing structures, offering cost savings, ease of management and scalability. Several companies have adopted cloud solutions in private environments, which allows them to reduce costs while still maintaining the ownership and management of their datacenters, mitigating some of the problems that are commonly encountered in Public Clouds such as multi-tenancy and security. Environments like these make continuous and intense use of Virtual Machines, taking advantage of the benefits of virtualization. Virtual machines can constantly migrate between physical servers, with different objectives, and may negatively impact the performance of the services deployed in them, and in the datacenter as a whole, however, migration is a feature that favors the management of Cloud environments. Therefore, constant monitoring of the environment is necessary, allocating the virtual machines in the best possible way in the physical servers, so that the objectives of the users of the datacenter are reached. The approach developed here, called VMPOS, performs the dynamic positioning of virtual machines, through a memetic algorithm that assists in the search for the best allocation of these machines, considering multiple objectives to be met simultaneously, in addition to applying resource overhead with different levels in physical servers, in order to obtain better performance for services considered critical. Thus, it is possible to configure the level of compromise of physical machines, providing flexibility to the datacenter and allowing more or less virtual machines to be allocated by physical machines according to the services that execute in each virtual machine. As a result, in comparisons with three other positioning approaches, it was possible to confirm that VMPOS reduced power consumption and increased server consolidation by applying resource overhead in the private cloud.