Pricing and Resource Allocation in Caching Services with Multiple Levels of QoS

(Research Seminar, February 05, 2003)

Kartik Hosanagar

Carnegie Mellon University

Abstract

Internet infrastructure is a key enabler of e-business. The infrastructure consists of backbone networks (such as UUNET and AT&T), access networks (such as AOL and Earthlink), Content Delivery Networks (CDNs - such as Akamai and Digital Island) and other cache operators. Together, all the players make up the digital supply chain for information goods. Caches provisioned by the access networks and the CDNs are the storage centers - the digital equivalent of warehouses. These caches store and deliver information from the edge of the network, bringing efficiency and responsiveness to content delivery on a global scale. The benefits of caching to content publishers, namely scalable content delivery, reduction in bandwidth costs and improvements in response times, are well recognized. Yet, caching has not been fully embraced by content publishers since its use can interfere with site personalization strategies, dynamic content creation and the potential loss of information about visitors to the site. Recent work on web caching has focused on the technological advances required to address these deficiencies. However, there has been no work on the managerial issues related to the design of incentive compatible caching services, appropriate pricing schemes and associated resource allocation issues that arise in operating a caching service. We propose a framework to enable an access network to provision vertically differentiated caching services that we refer to as "QoS Caching". We then develop analytic models to investigate optimal pricing and capacity allocation policies and discuss the managerial implications of the models for caching service providers.