Fog computing architectures are composed of a large number of machines distributed across a geographical area such as a city or a region. In this context it is important to support a quick startup of applications deployed in the for of docker containers. This thesis explores the reasons for slow deployment and identifies three improvement opportunities: (1) improving the Docker cache hit rate; (2) speed-up the image installation operation; and (3) accelerate the application boot phase after the creation of a container.Les architectures Fog computing sont composées d'un grand nombre de machines dispersées dans une zone géographique telle qu'une ville ou une région. Dans ce contexte il est important de permettre un démarrage rapide des applic...