Applications of Nonlinear Combinatorics in Optimizing Network Traffic

Abstract

Network traffic optimization is a critical aspect of ensuring efficient and reliable data transmission in complex communication systems. The graph approach, which is the foundation of most modern theories and optimization methods, is not sufficient when operating with a traditional toolkit of linear programming for the nonlinear and combinatorial nature of many networks. In this research, we focus on using submodular optimization and discrete Newton methods to improve the network performance as combinatorial optimization problems. The proposed methodology reduces the total end-to-end delay, increases the throughput, and also optimizes the time over which convergence occurs, which provides a reliable framework for controlling network traffic. By performing simulations on a graph-structured network, the two solutions as a one generated 35% lower average end-to-end delay, 25% higher overall throughput, and faster convergence compared to disparate approaches. From these results, we gain an understanding of the importance of handling nonlinear cost functions and the capacity of submodular optimization and DNM. As such, key issues such as scaling up to a large network and its ability to handle time-varying traffic remain limitations and thus require research. The results of this study hold valuable implications for network management allowing a timely route to re-organize the traffic load without overly expensive infrastructural evolutions. This research builds on important limitations of the current approaches to distill knowledge for further development in the context of network traffic shaping, which could be important for the development of 5G, IoT as well as numerous other up-and-coming concepts.