Social Welfare Implications of Different Cache Deployment Scenarios

Content delivery from distributed storage nodes within the network provides both an evolutionary and a revolutionary means to deal with the proliferation of content retrieval on the Internet. Regardless of whether the architecture redirects requests to nearby nodes that possess copies of a desired object or directly searches for the desired object by name, distributed storage plays a vital role in facilitating content delivery. Despite the important role played by distributed storage nodes in the current and future Internet, it is not yet obvious who should deploy them. Further, the social welfare implications of different deployment scenarios are not well understood. In this paper, we use a simple economic model to study the welfare implications of different cache deployment scenarios. We show that the level of caching supplied with transparent caching is inefficient and results in a deadweight loss. With a payment flow from publishers for content delivery, the level of caching supplied increases because caching service providers can extract some of the publisher surplus for themselves. Moreover, we show two deployment scenarios that maximize social welfare when a payment flow exists between the publisher and the caching service provider. In the first, eyeball networks provide the caching infrastructure through a federation or a third-party entity that serves as a transaction broker between publishers and various networks. In the second, eyeball networks pay third-party content delivery networks an amount, equivalent to the realized bandwidth and transit savings from caching, to deploy caches. Even though both deployment scenarios maximize welfare, we show that the latter deployment scenario is unlikely to be realized in practice because eyeball networks do not share in the resulting value created. This explains the emergence of caching federations and content distribution networks (CDNs) becoming brokers for eyeball network caching services