Caching and capacity allocation game in self-managed content provider networks

Abstract

In this paper we study the problem of distributed cache management in a self-managed network. We consider a set of autonomously operating edge routers, providing the wide-area connectivity for a local network. Each local network is equipped with a caching server, capable of storing copies of data objects. Objects can be either accessed from a special publisher server, or from any other cache that contains a copy. It is assumed that all data transmissions are carried trough links of limited bandwidth capacity. The problem consists of jointly deciding on the contents of all cache servers and setting appropriate transmission rates, so as the value of network utility is maximized. Due to the distributed nature of the considered system, we employ a game-theoretic analysis, and present how to compute efficiently solutions which constitute pure Nash equilibria. We introduce the concept of conditional equilibrium, which corresponds to a strategy of a game with restricted players' strategy space. Finally, an experimental study is presented, which consists of an application of the presented solution algorithm to existing wide-area network (Pionier optical network).