Agent-based Modeling and Simulation of Cooperative Content Distribution Networks

Content distribution networks (CDNs) have recently become the most adopted solution to efficiently delivery content through the Internet. To improve their performances, new techniques and schemes should be carefully designed and analyzed. This paper proposes the agent-based modeling and simulation of cooperative CDNs which are CDNs with surrogate servers clustered into cooperative groups. The modeling phase relies on statecharts-based agents and event-based interactions which ease the design of the active elements of a cooperative CDN. In particular, three CDN architectures have been modeled: conventional CDN, cooperative CDN with full managed clusters, and cooperative CDN with peer-to-peer-based clusters. Simulation, which is based on the Repast toolkit, allows for the evaluation of the average user perceived latency and cache hit ratio of the modeled architectures. The obtained results show that the proposed cooperative CDN architectures have the potential to be more efficient with respect to conventional CDN architectures.