CoAP Infrastructure for IoT

Abstract

The Internet of Things (IoT) can be seen as a large-scale network of billions of smart devices. Often IoT devices exchange data in small but numerous messages, which requires IoT services to be more scalable and reliable than ever. Traditional protocols that are known in the Web world does not fit well in the constrained environment that these devices operate in. Therefore many lightweight protocols specialized for the IoT have been studied, among which the Constrained Application Protocol (CoAP) stands out for its well-known REST paradigm and easy integration with existing Web. On the other hand, new paradigms such as Fog Computing emerges, attempting to avoid the centralized bottleneck in IoT services by moving computations to the edge of the network. Since a node of the Fog essentially belongs to relatively constrained environment, CoAP fits in well. Among the many attempts of building scalable and reliable systems, Erlang as a typical concurrency-oriented programming (COP) language has been battle tested in the telecom industry, which has similar requirements as the IoT. In order to explore the possibility of applying Erlang and COP in general to the IoT, this thesis presents an Erlang based CoAP server/client prototype ecoap with a flexible concurrency model that can scale up to an unconstrained environment like the Cloud and scale down to a constrained environment like an embedded platform. The flexibility of the presented server renders the same architecture applicable from Fog to Cloud. To evaluate its performance, the proposed server is compared with the mainstream CoAP implementation on an Amazon Web Service (AWS) Cloud instance and a Raspberry Pi 3, representing the unconstrained and constrained environment respectively. The ecoap server achieves comparable throughput, lower latency, and in general scales better than the other implementation in the Cloud and on the Raspberry Pi. The thesis yields positive results and demonstrates the value of the philosophy of Erlang in the IoT space.