Evaluation of Performance in Container Orchestration with Single-Board Computers in k3s Cluster

New technology, integrated tools, and intricate, reusable experiments in distributed computing infrastructures are needed to meet the requirements of modern scientific research. Above all, edge computing requires computational capacity for efficient data processing and analysis. In microservice and edge computing applications, single-board computers and container technologies have emerged as a new model for handling such intricate scientific problems. A single-board computer was constructed with one master and seven worker nodes in the lightweight Kubernetes k3s environment with NFS Server in the same network with an additional Raspberry Pi. A nginx server application was deployed in a Raspberry Pi k3s cluster, and the performance metrics were evaluated by scaling the pods up and down while applying load using external benchmarking software. The performance of the k3s cluster with the load balancer failed to handle transaction requests as the data size increased. It was found that the k3s nodes in the Raspberry Pi cluster failed due to frequently malfunctioning SD cards.