Evaluating Resource Consumption of Container Systems for Edge Computing

Sammanfattning

Edge computing is a steadily growing part of the world with more devices being implemented daily across the globe. Containerization is also becoming increasingly more popular in edgesystems due to being lightweight, providing consistency, and improving security. However, for expansion to continue effectively at current pace, knowledge about container systems resource and power usage is vital to ensure edge systems are sustainable and devices maintain long evity and efficiency throughout. This study investigates the resource- and power usage of multiple container systems on single board computers through a quasi-experiment. The experiment is conducted by analysing data gathered from three different single board computers running four variations of container systems. The data focuses on CPU and memory usage through top, as well as power consumption through RAPL with the help of Powerletrics. The results showed differences between single board computers memory usage running the same container system, as well as consistent patterns between container systems in terms of power usage as well as CPU- and memory utilization. An attempt to explain these is done, however due to data being limited it is not possible to narrow down further than hardware, architecture, and OS all possibly affecting gathered data. Results are also compared to related works, where inconsistencies are highlighted across works, which is presumed to be for the same reason. This is done to highlight the impact hardware and system design has on the resource usage of container systems, with a deeper analysis being proposed as future work.