• constrained application protocol
• message queueing telemetry transport

These days many enterprise systems involve applications collecting data from various locations, analyze this data, and based on the analysis possibly issue an action. Moving the data uncompressed over the network requires significant network bandwidth, so making partial computations along the path and moving only a small set of data to the cloud is a potential solution to mitigate this problem. As a result, it is found to be beneficial to automatically deploy, run, and monitor machines that provide many services near to the place where data is acquired. To this aim Fog computing is envisioned, an intermediate computing layer to be deployed between the sensor nodes and the cloud to collect and analyze the data directly in the proximity of the cyber-physical systems. One such platform is KubeEdge, a virtual machine orchestration software based on Kubernetes. It enables the creation, deployment, and management of containers on Fog nodes. Recently there have been many efforts to enrich this platform to have many capabilities that we enjoy in the cloud computing platform: automatic deployment, scalability, scheduling, live migration, and so on. These efforts include enabling the Fog units to communicate with each other, with the sensing units, and with computers that are running in the cloud. Due to the nature of sensor nodes, these communication protocols should be able to fulfill many requirements such as less power consumption, scalability, resiliency to network failures, and interoperability with existing systems. Considering the constrained nature of sensor nodes many low power communication protocols are being used. Among these low power communication protocols are MQTT, which works in a publish-subscribe manner, and CoAP, which works in a request-reply manner. In Kubeedge, MQTT is already implemented using Mosquitto Broker, but CoAP communication is not implemented. This thesis enriches the KubeEdge platform by incorporating communication functionalities with CoAP based sensor nodes. To this end, we took sample sensors and actuators which communicate via CoAP protocol and developed mappers to interface these sensors with KubeEdge. These mappers act both as a client and a server, deployed on the edge nodes, and translates request/response communications to publish/subscribe communications to synchronize the sensor state with its cloud and edge representation. Besides a CoAP resource directory draft implementation and integration with Kubeedge is performed to illustrate how to automatically manage the resource discovery of CoAP server mappers via Kubeedge. Furthermore, the performance of the mappers and the resource directory implementation is measured through emulations. The resource directory is developed in such a way that it allows to be installed in a stand-alone or distributed manner. The emulation experiment results depict an interesting trade off between consistency and response latency of different installation setups of the resource directory.