Comparative analysis between four models of IoT gateway selection

Abstract

The “Internet of things” (IoT) is a leading sector of technology where usual analog devices can be turned into smart devices with embedded systems and sensors. These devices can communicate with each other without requiring any human to human or human to computer interaction and automate various aspects of human lives. With the growing application of IoT some of the crucial challenges faced in this sector are maximizing throughput, minimizing energy consumption of the gateways and balancing loads among the gateways efficiently. Not to mention providing high throughput and low energy consumption at the same time is a contradictory concept and the selection models have to reach the most optimum trade-o↵ point to o↵er better performance. There are various models that have been proposed for IoT gateway selection, each of which comes with various pros and cons. Hence we decided to conduct a comparative analysis between a few of the most innovative and promising Gateway Selection models to find out which of these models are relatively more e↵ective to tackle the above-mentioned challenges in IoT. For our research, Game Theory Gateway Selection (GTGWS), Taxi-Sharing, Floyd-Warshall & Minimax (FWM) Algorithm and Evolved Reliability and Traffic-aware Gateway Selection (ERTGS) model were studied. To ensure the credibility of this analysis the models are tested under two distinguished network conditions - di↵erent ‘demand’ of the end devices & di↵erent ‘number’ of end devices connected to the gateway. Under the given scenarios, various data were obtained. For instance, given a least congested network, bandwidth usage by these four models were: ERTGS utilizes 30.464%, Taxi-Sharing utilizes 15.25%, FWM utilizes 12.5% and GTGWS uses 21.68% of the total bandwidth. Various results like these were obtained for other evaluation criteria such as load di↵erence and energy consumption; after careful analysis it was found that none of the four models o↵ers optimal solutions to all of the challenges; and that di↵erent models are better suited for di↵erent network priorities which are discussed in details in this paper.