Benchmarking Kubernetes based Edge Server in Embedded Environment

초록

Recently, there are increasing attempts to apply deep learning and machine learning to embedded boards in edge computing environments. However, due to the limited environment of embedded boards, problems such as virtualization, distributed control, and message communication are occurring. This paper builds a Kubernetes-based edge server in various embedded environments and benchmarks the performance of the workload. The experimental results show that the network bandwidths at Nano, Xavier, and NUC measured by iperf3 as Network measurement program are 232Mb/s, 257Mb/s, 271Mb/s and the response time of image/video transmission at Nano, Xavier, and NUC measured by Socket program in container environment are 49.98ms, 50ms, 43ms, respectively.