Efficient resource management and workload allocation in fog-cloud computing paradigm in IoT using learning classifier systems

With the rapid growth in network-connected computing devices, the Internet of Things (IoT) has progressed in terms of size and speed. Subsequently, the amount of produced data and computation loads has increased dramatically. A solution to handle this huge volume of workloads is cloud computing in which a considerable delay exists in the processing load and this has remained a concern in the field of distributed computing networks. Processing workloads at the edge of the network can reduce the response time while at the same time imposing energy constraints by bringing the task of load processing from data centers, which are supplied by electrical energy sources, to the network edges which are only supported by limited energies of batteries. Therefore, workloads need to be distributed evenly between the clouds and the edges of the network. In this paper, two methods based on XCS learning classifier systems (LCS), namely, XCS and BCM-XCS, are proposed to balance the power consumption at the edge of the network and to reduce delays in the processing of workloads. The results of our experiments are indicative of the superiority of BCM-XCS over the basic XCS-based method. The proposed methods distribute the workloads in a way that the delay in their processing and the communication delay between the cloud and fog nodes are both minimized. In addition to considerable advantages in controlling the fluctuations of the processing delay, the proposed methods can simultaneously reduce the processing delay by 42% by using a moderate power consumption at the edge of the network. The proposed methods can also recharge the renewable batteries used at the edge of the network about 18 percent more than the best state-of-the-art method.