A Network-Based Fog Computing (NBFC) Model of the IoT

In the IoT (Internet of Things), a huge amount of electric energy is consumed to transmit a tremendous amount of sensor data and to execute various types of resource-consuming applications like AI which take usage of the sensor data on servers in clouds. In this paper, we consider the FC (Fog Computing) model of the IoT, where some application processes of sensor data are supported by fog nodes. A fog node obtains output data by processing input data from a preceding fog node and then sends the output data to a succeeding fog node. In the TBFC (Tree-Based FC) model, fog nodes are tree-structured where a child-to-parent relation shows node -to-succeeding node relation in order to reduce the total energy consumption of the IoT. Here, each node can communicate with the parent node and the child nodes. If some node gets faulty and performance bottleneck, the tree structure has to be changed to recover from the faults and increase the performance. In order to make the FC model more flexible, we newly propose an NBFC (Network-Based FC) model where a succeeding node is found each time a node finishes the computation on received data. We also propose a pair of L1 and L2 algorithms to find a succeeding node for each node. By using the algorithms, a path from an edge node to a server node is obtained. In the L1 and L2 algorithms, a node is selected to be a succeeding node where the energy consumed by each process and the total energy consumption of the node are the smallest is selected, respectively. In the evaluation, the total energy consumption of nodes in an edge-to-server path obtained in the L1 algorithm is smaller than the L2 algorithm.