Analyzing the Power Consumption in Cloud-Native 5/6G Ecosystems

The Fifth generation(5G) technology moves a step forward from the previous generations by natively supporting vertical applications with real time and low latency requirements. However, the resultant increase of computing resources will cause a consequent increase in the energy requirement and an impact on the carbon footprint. To this end, it is of paramount importance to involve all the stakeholders leveraging on the 5/6G ecosystem, by making them aware of the impact of their applications and network services on the underlying hardware infrastructure and promote, through economic incentives, a new value chain that puts energy efficiency in the spotlight. To achieve this goal, monitoring tools are required to measure the performance and resource consumption at the infrastructure level and map it to the virtual images on top of it to infer the consumption ascribable to a vertical application or a network function. Specifically, this paper analyzes the resource and power consumption of a Kubernetes container, focusing on the breakdown of CPU usage, wakeups and power consumed by the individual kernel processes in the presence of traffic varying in load and composition. Experimental results compare an application by itself with its containerized counterpart and show that the latter requires much more time and effort to process data, and uses more resources compared to a stand-alone application.