The Internet is currently mostly exploited as a means to perform massive digital content distribution. Such a usage profile was not specifically taken into account while initially designing the architecture of the network: as a matter of fact, the Internet was instead conceived around the concept of host-to-host communications between two remote machines. To solve this problem, Content-Delivery Networks (CDNs) are currently used as a well-established technology to serve content-driven demands through an infrastructure that is not tailored for that purpose. On the other hand, the novel paradigm of Content-Centric Networking (CCN) aims at filling the gap of this misalignment by changing the network-layer protocols, solving the content-distribution problem at its root. In this paper, we formulate novel optimization models to analyze the performance gains that CDN and CCN can achieve, by reducing the total amount of traffic exchanged through the network. We tackle this problem by adopting a time-varying content popularity evolution model that accurately represents the dynamic behavior of users. We discover that, in most of the cases, CDN reduces the overall traffic exchanged between network nodes, leading to better performance, whereas CCN should instead be preferred in those scenarios where CDN cannot quickly react to popularity evolution. On top of that, we show that very limited benefits can be obtained by changing the cache replacement algorithms. Finally, all our key findings are confirmed by extensive simulation campaigns that further complement this work.
Submitted for publication to IEEE/ACM Transactions on Networking.