Aprimorando a elasticidade de aplicações de banco de dados utilizando virtualização em nível de sistema operacional

Abstract

In order to maintain a service running within the acceptable Qos levels, cloud providers should adapt resource provision to handle workload fluctuations. In case of under provisioned resources service performance tend to be affected while over provisioned resources reduce providers profit. To cope with this scenario several studies present elastic solutions for resource allocation in virtualization environments. However, in traditional virtualization environments, the elasticity is limitedby the limits established during initialization of virtual machines, not allowing the addition of resources in runtime. Instead, OS-level virtualization is a new approach that allows the manipulation of all available resources shared by a host at runtime. Therefore, this study evaluates the impact of applying the dynamic resource allocation strategy proposed by Dawoud et al., firstly developed for traditional virtualization environments, to OS-level virtualization. To evaluate our approach, we use database applications because they are widely used in virtualized environments, and their workloads use processor, memory, and network. The results show that our approach has a more efficient use of available resources, resulting in a positive impact on performance metrics.