Comparative experimental analysis of the quality-of-service and energy-efficiency of VMs and containers' consolidation for cloud applications

Abstract

Abstract-The consolidation of services is a widely accepted technique for IaaS Cloud providers to reducing energy consumption and improving the utilization of their resources. This technique is based on distributing all services in the minimum amount of servers. This way, the overall energy consumption of the datacenter is reduced, as less servers are needed to be active. Traditionally, research has focused on strategies for consolidation of Virtual Machines (VMs), but containers are changing the landscape of Cloud services. Containers are expected to optimize the consolidation of services by reducing the amount of needed resources, thus allocating more services using less servers. However, while multiple research works have been produced in the Energy Efficiency (EE) achieved through consolidation of VMs, there is yet no experimental work on how consolidation of containers affects EE, when assuming a given Quality-of-Service (QoS) to the user. In this paper we show an experimental analysis on the effects of consolidation of containers in the QoS and EE, compared to the consolidation of VMs. We demonstrate that the consolidation of containers is indeed more optimal than the one of VMs, both in terms of QoS and EE. Consecutively, we analyze how the degradation of the service is produced both in QoS and EE, and we show how QoS is the variable which is more affected by consolidation. This work provides the necessary scientific background on consolidation of two widely used virtualization technologies, and we believe it is useful for future works on the optimization of resources in datacenters.

Energy efficiency in cloud computing data centers: a survey on software technologies

“…Docker, when configured to a maximum latency of 3000ms, can operate up to 21% more services than KVM. Docker provides this service in this setup while consuming 11.33% less energy than KVM [39].…”

Sobre a simulação de algoritmos de escalonamento cientes de energia renovável

“…Por um lado, a capacidade de mapear os recursos computacionais físicos em recursos virtuais permite consolidar mais aplicac ¸ões em um único servidor. Por outro, se torna mais difícil monitorar qual é a real utilizac ¸ão de hardware dos recursos virtuais e seu consequente consumo elétrico [Cuadrado-Cordero et al 2017].…”

A study on the evaluation of HPC microservices in containerized environment

“…The study by Morabito 35 compares the power consumption of container and VM and shows that both types of virtualization have similar power consumption for idle situations or for CPU/memory operations, but containers consume less power for network‐intensive operations. Cuadrado‐Cordero et al 36 compared the QoS and energy performance of Docker containers and KVM for different services. These experimental results show that Docker allows more services to run compared with KVM.…”