Toward Delay-Based Utility Maximization: Modeling and Implementation in an SDWN Platform

Packet (link) scheduling algorithms in wireless networks should be able to maximize the networks' transmission capability while maintaining an acceptable delay and fairness performance. Maximum-weight (MW) policies based on the delay index are throughput-optimal in a wireless network. However, sampling the real-time head-of-line (HOL) packet statistics is still challenging. We model the delay as a function of the arrival rate and service rate, and this model enables the scheduling algorithms to obtain the packet delay without high-frequency sampling. Then, we model the delay-related performance problems in everyday life as a utility-maximization problem. By utilizing the Lyapunov optimization, we transform the initial problem into a queue stability problem and design an MW algorithm to solve it. The solution can ensure deterministic long-term average delay guarantees. Moreover, its throughput utility differs from the optimally fair value by an amount that is inversely proportional to the delay guarantee. We evaluate the theoretical performance via simulations. We also build a real-world software-defined wireless network platform and implement our algorithm on it to evaluate the algorithm's actual performance. The experimental results show that our method outperforms the state-of-the-art algorithm in terms of throughput by up to 13.83%, and it outperforms the default first-in, first-out (FIFO) method by up to 40.41%.