Towards optimal content replication and request routing in content delivery networks

Cooperative content replication and request routing (C²R³) has emerged as a promising technique to enhance the efficiency of content delivery networks (CDN). Most existing approaches to C²R³ focus on efficient bandwidth usage and assume a hierarchical CDN architecture targeted towards the delivery of specific content types (e.g., video). Therefore, C²R³ problem of covering the broad range of content types with minimum content access delay in a general CDN architecture has attracted little attention. As a potential solution to C²R³, cooperative web caching techniques have become mature. However, these techniques were designed to improve performance indicators tailored to web contents only (i.e., hit rate and byte hit rate). Arguably, improving such indicators does not necessarily lead to optimal access delay especially when the current trend of user-generated contents with diverse popularities and sizes are taken into account. In this paper, we formulate C²R³ as an optimization problem with the objective of minimizing content access delay in a general CDN architecture. A new performance indicator is introduced, and two popularity-based cooperative algorithms are proposed to approach the NP-hard C²R³ problem. Under broad ranges of cache size and popularity distribution parameters, we compare the proposed methods with a cooperative recency-based web caching method. Our simulation results show that the popularity-based methods outperform the recency-based method, and demonstrate close to optimal performance in representative scenarios of real-world situations.