An Empirical Evaluation of the Constrained Application Protocol in ARM-Based Platforms

Single-Board Computers (SBCs) have been expe-riencing a surge in popularity, driving innovation in the de-velopment of a wide range of automated systems. SBCs are convenient since they are affordable, versatile, easy to use, and have low power consumption. Furthermore, the world of the Internet of Things (IoT) is dramatically expanding with billions of devices deployed and connected to the Internet across the earth. In order to facilitate the communication of these devices, several network protocols have been proposed by the community, where CoAP seems to be one of the more widely spread. This paper benchmarks the response time of common CoAP requests (PUT/GET) when varying several parameters such as the payload size, block size, transport protocol, security options, network technologies, and development tools over three Raspberry Pi Models: Raspberry Pi 3 Model B (RPi 3B), Raspberry Pi 3 Model B+ (RPi 3B+), and Raspberry Pi 4 Model B (RPi 4B). In these experiments, the RPi 4B did notably better than the other two devices under test, which had alike performance. When security is of concern, adding options for PSK-based DTLS decreased the performance by up to 12%. Using UDP or TCP as a transport protocol did not make a significant difference. Regarding the development tools, libcoap notably outperformed aiocoap. However, the steep learning curve of the former may be a limitation for its adoption. With this paper, the authors intend to guide users in the selection of SBCs, development tools, and adequate parameters of CoAP for their use cases when deploying IoT proiects,