The need for reliable, responsive, and rapidly updatable applications that run in the cloud and at the network edge has been increasing in recent years. To achieve these goals, containers are being used increasingly, as they enable efficient, portable, and consistent application execution. However, managing a large number of containers requires the use of orchestration platforms such as Kubernetes and Docker Swarm. This master’s thesis focuses on the energy consumption of orchestration systems. Based on the most common types of system workloads, we designed test cases that reflect the differences between
the two platforms. We analyze the differences between running applications without an orchestration platform, running on a single node, and running on a worker node. The experiments are carried out on devices with varying capabilities. The result of the thesis is a functional system for testing energy consumption and a set of guidelines that highlight the differences between the functionality and energy usage of orchestration systems.
|
