An agent-based approach for the design and analysis of content delivery networks

Content delivery networks (CDNs) are currently the most exploited worldwide solution for reliable, scalable and cost-effective delivery of content over the Internet. The development and deployment of new services and components in CDNs should be carried out carefully as they involve huge investments from the CDN providers’ perspective. It is therefore necessary to prototype, monitor, and predict the behavior of new CDN services/components in a controlled simulated environment, before and after their final release. As agent-oriented methods, particularly agent-based modeling and simulation (ABMS), have demonstrated high effectiveness for the analysis and design of modern complex distributed systems in a wide range of different application domains, this paper proposes an effective approach based on the agent-oriented ELDAMeth methodology for the design and analysis of CDNs. Specifically, an agent-based extensible CDN framework provides high-level programming abstractions and tools that fully support the simulation of different CDN architectures and mechanisms and allow for automatic evaluation of three main CDN performance indices: average user perceived latency, cache hit ratio and utility. The framework currently makes it available agent models for conventional, cooperative and clustered CDN architectures that have been simulated according to a reference simulation scenario. Results clearly show that clustered architectures have the potential to outperform conventional and cooperative architectures on which real CDNs are currently based and that the proposed agent-based framework is effective in modeling and analyzing CDNs.

► An agent-oriented simulation framework for the design and analysis of content delivery networks.
► An agent-based modeling and simulation methodology for the lifecycle of CDNs.
► Clustered architectures can outperform currently available conventional and cooperative CDNs.