Algoritmo de seleção de políticas de migração de máquinas virtuais em função de diferentes demandas de recursos

Abstract

This study suggests the application of an algorithm to choose: i) the virtual machine migration approach, whether static or dynamic; ii) the strategy for selecting the destination cloudlet in the virtual machine, which can be based on the lowest latency between the source cloudlet and one of the destination cloudlets, the shortest distance between the user and one of the destination cloudlets, or the cloudlet connected to the nearest user access point; and iii) the virtual machine migration policy, which can be pre-copy or post-copy, depending on system parameters such as signal quality, bandwidth, latency, and virtual machine size. The proposed algorithm performs the migration of the virtual machine while the user moves, keeping the internet connection active. This results in the reduction of the virtual machine migration time, the virtual machine downtime, and the volume of data migrated from the virtual machine, through the dynamic selection of the migration approach, strategy, and policy, according to the quality of the input parameters. The main metrics for evaluation are the total migration time of the virtual machine, the downtime of the virtual machine, and the total volume of data migrated from the virtual machine. These metrics are essential to maintain the user’s internet connection during their movement in a fog computing scenario. The results obtained with this algorithm surpassed those of other algorithms, presenting a superior efficiency of over 6%.