Optimal Weight Design for Decentralized Consensus-Based Large-Scale Data Analytics over Internet of Things

We are witnessing an increasing demand for smarter systems requiring more sensory data for decision makings. The main data sources are connected objects to the network forming the Internet of things (IoT). Processing such huge amount of data, coming back from IoT devices forces big data challenges in both storage and processing. Different solutions exist for data storage and analysis over IoT including cloud, Edge and Fog platforms that mainly focus on decentralization of the data. To process the decentralized data with no need to a center, optimized distributed decision making algorithms are required. In this paper, we study one of the promising distributed algorithms named consensus to make a decision over decentralized data generated by IoT devices. It is assumed that the data sources are capable to analyze limited portion of data while share their decisions and partially of their data with neighbor sources. Sharing the information, the network objects combine their decisions with neighbor sources and reach to a consensus using iterative optimization algorithms. We propose an optimal weight design for IoT devices over the practical networks. We assume that the data transmission is noisy and the weights are chosen so that the network cost function is minimized. Using the proposed weighting the network reaches a faster convergence while the resource consumption decreases.