Dynamic Big-Data Broadcast in Fat-Tree Data Center Networks With Mobile IoT Devices

In this paper, we study the problem of throughput and delay-optimal dynamic big-data broadcast in fat-tree data center networks (DCNs) in the presence of mobile Internet-of-Things (IoT) devices, where one of the IoT devices acts as a source node. In existing literature, researchers studied that a balanced traffic distribution in DCNs is a NP-hard problem. With the integration of heterogeneous IoT devices in DCNs, the difficulty in achieving a balanced traffic distribution increases significantly. Hence, there is a need to design a throughput and delay-optimal big-data broadcast scheme in DCNs in the presence of IoT devices. In this paper, we propose a dynamic big-data broadcasting scheme, named D2B, using a single-leader-multiple-follower Stackelberg game for solving the aforementioned problem. Here, each switch acts as the leader, and the IoT devices act as the followers. We consider that the source node broadcasts the generated data in real time. We represent bandwidth distribution as a pseudo-Cournot competition, where each follower decides the optimal downloading bandwidth. The existence of the generalized Nash–Stackelberg equilibrium for D2B is evaluated theoretically. We observed that using D2B, the network throughput increases by $\text{55.32}\%$, while ensuring at least $\text{33}\%$ increase in the average bandwidth allocation per IoT device, and the overall delay in broadcasting is reduced.