A Shallow Deep Neural Network for Selection of Migration Candidate Virtual Machines to Reduce Energy Consumption

Abstract

In recent years, the widespread growth of cloud computing has surprisingly increased the energy consumption in data centers. In this regard, employing energy reduction techniques is changed to one of the prominent challenges for cloud service providers and includes both dynamic and static techniques. Although by utilizing static techniques along with creating data centers energy consumption is relatively reduced, the rapid growth of cloud computing due to the increasing demands of users for these resources has changed energy consumption to a potential challenge. Utilizing dynamic energy reduction techniques which can be possible through the integration of the virtual machine into at least one physical server can be considered as an effective solution to this problem. This is done through live virtual machine migration and selecting the migration candidate virtual machine is a key step in this technique. In this paper, the combination of Convolutional Neural Network (CNN) and Gated Recurrent Unit (GRU) is used to choose the appropriate migration candidate virtual machine which leads to the diagnosis of whether a virtual machine is sensitive to latency or not. The proposed model was validated on the workload of Microsoft Azure virtual machines as a dataset. According to the empirical results, the proposed model has higher classification accuracy compared to other existing models for selecting the migration candidate virtual machines.