摘要： The rapid increase in the number of Internet connected devices has placed a high level of demand on both the network bandwidth and the processing power currently available to the centralized physical datacentres that embody the ‘Cloud’. To alleviate such a demand and relieve the network capacity, the Edge Computing paradigm was recently proposed. Such a paradigm evangelizes the processing of data locally, closer to users and so avoiding the slow communication of data to centralized datacentres. The realization of such a paradigm requires the evaluation of smart applications on various potential energy efficient devices by understanding their processing and storage limits, while also looking for efficient methods to improve their capabilities. In this paper, we develop and evaluate an end to end smart office application on a representative Edge device, the latest Raspberry-Pi, while utilizing existing Cloud on Demand services connected through an Android application and Amazon Alexa Skill. The developed solution monitors various environmental conditions and is able to recognize users using facial recognition. We will evaluate the requirements of such an application on an Edge device and explore methods to reduce the data stored and processed, while evaluating the impact this has on the detection rate. In particular, we evaluate the effectiveness of increasing the image compression levels by measuring the level of compression used versus the time to create trained data and the face detection rate it produces. Our experimentation shows favourable results for the Edge system implementation, while also supporting the possibility of the hybridisation of Cloud and local processing to achieve complex tasks while minimising network use.