An algorithm to restructure the Semi-Homogeneous Tree-Based Fog Computing (SHTBFC) model to reduce the energy consumption

In the IoT (Internet of Things), a large amount of electric energy is consumed since a tremendous volume of sensor data is transmitted from devices to servers in networks and large computation resources are spent to execute sensor applications on servers. In order to realize the energy-efficient IoT, we discuss an $SHTBFC$ (Semi-Homogeneous Tree-Based Fog Computing) model where types of servers and devices are interconnected in a tree of homogeneous fog nodes. Here, a root stands for a server and a leaf indicates a device. Thus, all the nodes are not homogeneous in an SHTBFC model. Each fog node generates output data by processing input data from the child fog nodes and passes the output data to the parent node. If some fog node $f$ does not behave correctly due to faults, the descendant fog nodes of $f$ cannot communicate with the ancestor nodes. In order to recover from the faults, improve the performance, and reduce the energy consumption, we newly propose a $TR$ (Tree-Restructuring) algorithm to restructure an SHTBFC tree to an SHTBFC tree which consumes smaller energy. An objective fog node is first taken, which consumes the largest energy in an $SHTBFC$ tree. Then, one of three $TR$ operations, $M$ (migrate), $R$ (replicate), and $X$ (expand) operations is executed on the objective node. The energy consumption of an application depends on the number of application processes. The more processes an application is composed of, the more fog nodes are included in an SHTBFC tree obtained by the $TR$ algorithm. In the evaluation, we show the number of application processes where an application consumes the smallest energy.