fCDN: A Flexible and Efficient CDN Infrastructure without DNS Redirection or Content Reflection

TL;DR

This paper introduces fCDN, a novel CDN architecture that eliminates DNS redirection and content reflection by leveraging ICN routing, improving resource utilization and reducing path lengths in content distribution networks.

Contribution

The paper proposes a flexible CDN architecture using ICN-based routing and a new resource placement algorithm, Swing, to optimize edge-to-edge content delivery without DNS or reflection.

Findings

fCDN reduces edge-to-edge path length

fCDN improves network resource utilization

Swing algorithm outperforms traditional placement algorithms

Abstract

Flexible and efficient CDNs are critical to facilitate content distribution in 5G+ architectures. Current CDNs suffer from inefficient request mapping based on DNS redirection, and inefficient content distribution from origin to edge servers, through content reflection. We proposes a novel, flexible CDN architecture that removes the need for DNS-based mapping and content reflection. Instead, requests to/from the CDN are treated as service transactions in the network, which utilises a routing function embraced from emerging research in Information-Centric Networks (ICN) to route edge-to-edge transactions to the true nearest service point. The same function is utilized to establish path-based flows over a fast forwarding substrate; thereby, eliminating the need for IP routing between service points within a single domain, and potentially at peering points with other domains. We model our architecture and formulate the resource placement problem as a variance of the K-center problem. To address the problem, we propose a greedy algorithm, Swing, that balances the placement of service points between highly and poorly connected nodes. We evaluate the efficiency of our architecture in utilising the CDN and network resources through Monte Carlo simulations that explore a range of K values. Moreover, we compare the goodness of the placement solutions provided by Swing with those provided by Largest First and Closest First Algorithms. Evaluation results show the superiority of our fCDN solution in reducing the edge-to-edge path length and the required network resources.