On Optimal Two-Sided Pricing of Congested Networks

Abstract

Internet Access Providers (APs) have built massive network platforms by which end-users and Content Providers (CPs) can connect and transmit data to each other. Traditionally, APs adopt one-sided pricing schemes and obtain revenues mainly from end-users. With the fast development of data-intensive services, e.g., online video streaming and cloud-based applications, Internet traffic has been growing rapidly. To sustain the traffic growth and enhance user experiences, APs have to upgrade network infrastructures and expand capacities; however, they feel that the revenues from end-users are insufficient to recoup the corresponding costs. Consequently, some APs, e.g., Comcast and AT&T, have recently shifted towards two-sided pricing schemes, i.e., they start to impose termination fees on CPs' data traffic in addition to charging end-users. Although some previous work has studied the economics of two-sided pricing in network markets, network congestion and its impacts on the utilities of different parties were often overlooked. However, the explosive traffic growth has caused severe congestion in many regional and global networks, especially during peak hours, which degrades end-users' experiences and reduces their data demand. This will strongly affect the profits of APs and the utilities of end-users and CPs. For optimizing individual and social utilities, APs and regulators need to reflect the design of pricing strategies and regulatory policies accordingly. So far, little is known about 1) the optimal two-sided pricing structure in a congested network and its changes under varying network environments, e.g., capacities of APs and congestion sensitivities of users, and 2) potential regulations on two-sided pricing for protecting social welfare from monopolistic providers. To address these questions, one challenge is to accurately capture endogenous congestion in networks. Although the level of congestion is influenced by network throughput, the users' traffic demand and throughput are also influenced by network congestion. It is crucial to capture this endogenous congestion so as to faithfully characterize the impacts of two-sided pricing in congested networks. In this work, we propose a novel model of a two-sided congested network built by an AP. We model network congestion as a function of AP's capacity and network throughput, which is also a function of the congestion level. We use different forms of the functions to capture congestion metric based on different service models, e.g., M/M/1 queue or capacity sharing, and user traffic based on different data types, e.g., online video or text. We characterize users' population and traffic demand under pricing and congestion parameters and derive an endogenous system congestion under an equilibrium. Based on the equilibrium model, we explore the structures of two-sided pricing which optimize the AP's profit and social welfare. We analyze the sensitivities of the optimal pricing under varying model parameters, .e.g., the capacity of the AP and congestion sensitivity of users. By comparing the two types of optimal pricing, we derive regulatory implications from the perspective of social welfare. Besides, we also evaluate the incentives of the AP and regulators to adopt the two-sided pricing instead of the traditional one-sided pricing that only charges on the user side.