Abstract:
Fog computing extends the cloud services to the edge of the network by taking advantage of edge devices that have sufficient IoT resources (i.e, storage, compute, and bandwidth). "A cloud closer to the edge" has been proved as a promising solution for avoiding unbearable latency and network capacity saturation with the proliferation of IoT end-devices. Lately, researchers have noticed the impact of cloud-fog cooperation on the performance of the network in terms of latency, network capacity and security. While the cloud could handle heavy-weight delay-tolerant tasks, the fog becomes in charge of all light-weight delay-sensitive tasks.In such integrated networks, resource management becomes a key challenge that must be addressed effectively. Moreover, researchers have been preferring the clustered network topology for the fog layer over the flat one. In this thesis we aim to design and study two different resource management variations for fog network: flat and a clustered variations respectively. Both variations are formalized as an optimization problem in order to maximize the IoT tasks to fog assignments while satisfying not only the resources requirements of the issued tasks, but its QoS requirements as well. The comparison of both variations shows that the flat approach gives better results in terms of overall and fog delay when increasing the number of clusters in the topology, while the clustered approach results in lower number of tasks being rejected. Moreover, a baseline approach is presented as well to evaluate our proposed approaches.