Optimal Purchasing Strategies and Market Equilibrium in Multi-resource Internet Access Markets
Abstract:
Content delivery services, especially video streaming, have taken a significant portion of the Internet traffic. In this era of contents, the Quality of Service (QoS) perceived by end-users should be among the primary concerns of content providers (CPs), because metrics representing poor service quality such as long loading time and low available bitrate could severely reduce a user's willingness to watch a video online. As service quality deteriorates due to the network congestion at bottleneck links, CPs have strong incentives to purchase additional access resources from Internet service providers (ISPs) that could address the issues. For example, CPs may negotiate premium peering agreements with the ISP or deploy distributed caches. Although either bandwidth or cache resources can improve the QoS of content delivery, less is known for the scenario of deploying a hybrid of premium peering and cache resources. CPs can achieve better QoS by acquiring more access resources, but they also induce additional resource expenses. In this work, we study the optimal strategies for CPs to procure different types of access resources such as peering bandwidth and cache from the ISP. We also study the market equilibrium in the last-mile access markets, at which the demand and supply are balanced.
Firstly, we propose a novel capacity-driven cache policy named Weighted Fair Caching (WFC). It is centered on enabling explicitly controllable allocation of cache resources in a work-conserving manner, which is not feasible under existing policies such as least recently used (LRU) and TTL-based caching. As a result, the cache owner is able to explicitly allocate both bandwidth and cache resources among competing customers. Secondly, we study how a CP determines an optimal purchasing strategy in order to maximize its individual utility, that is, the optimal quantities of bandwidth and cache resources to purchase. We formulate the problem based on a detailed CP-ISP-user model in the context of content delivery. Thirdly, we take a unified view of access resources available for purchasing from the ISPs in the last-mile access markets. In a multi-resource market, CPs intend to maximize their individual utilities, while the regulators may be concerned about the social welfare. We are interested in exploring the existence and uniqueness of an "optimal" market equilibrium, at which different optimization goals could all be fulfilled.
Our work provides insights for the thriving market of content delivery services, where CPs are willing to provide better experience for their users at a reasonable cost.