Remote robot control from Docker

Sammanfattning: Currently, we witness a new phase of digitization which is fueled by the development of Internet-connected smart sensors (Internet of Things - IoT). Also, about the processing of large data volumes that they create using Big Data analytics, leveraging the compute resources from the Cloud and Edge-based systems. For data exchange in the IoT world, typically lightweight communication protocols such as Message Queuing Telemetry Transport (MQTT) are used which are based on publisher/subscriber communication pattern where a broker mediates data among interested parties. In order to provide reliable communication, MQTT provides different Quality of Service (QoS) mechanisms. MQTT publishers, subscribers and brokers can be deployed inside containers on virtualized infrastructure to facilitate large-scale virtualized compute frameworks from the cloud for scalable data analytics. However, each docker containers requires a specific amount of resources to provide the required response time.In this thesis, we evaluate the impact of resource sharing due to the virtualized deployment of MQTT components on latency and response time of IoT applications. We deploy a testbed of Arduino and Raspberry Pi devices that host MQTT clients to pull sensor data towards MQTT clients inside the cloud. We also evaluate the impact of different QoS levels at the MQTT protocol on latency. Our results indicate that proper resource allocation and QoS parametrization is important for maintaining low and stable latency.