LI Zhe1,2, SIMON Gwendal1,2
Article de revue avec comité de lecture
IEEE transactions on network and service management, september 2013, vol. 10, n° 3, pp. 300-311
The exploding HD video streaming traffic calls for deploying content servers deeper inside network operators' infrastructures. Telco-CDN are new content distribution services that are managed by Internet Service Providers (ISP). Since the network operator controls both the infrastructure and the content delivery overlay, it is in a position to engineer telco-CDN so that networking resources are optimally utilized. In this paper, we show the following two findings: 1. it is possible to implement an efficient algorithm for the placement of video chunks into a telco-CDN. We present an algorithm, which is based on a genetic algorithm implemented on the MapReduce framework. We show that, for a national VoD service, computing a quasi- optimal placement is possible. 2. such push strategy makes sense because it allows to actually take into account fine-grain traffic management strategies on the underlying infrastructure. Our proposal re-opens the debate about the relevance of such "push" approach (where the manager of telco-CDN proactively pushes video content into servers) versus the traditional caching approach (where the content is pulled to the servers from requests of clients). Our proposal of a quasi-optimal tracker enables fair comparisons between both approaches for most traffic engineering policies. We illustrate the interest of our proposal in the context of a major European Telco-CDN with real traces from a popular Video-on-Demand (VoD) service. Our experimental results show that, given a perfect algorithm for predicting user preferences, our placement algorithm is able to keep aligned with LRU caching in terms of the traditional hit- ratio, but the workload on some troubled links (e.g., over-used links) in a push-based strategy is significantly alleviated.
1 : RSM(TB) - Dépt. Réseaux, Sécurité et Multimédia (Institut Mines-Télécom-Télécom Bretagne-UEB)
2 : IRISA(TB) - Institut de recherche en informatique et systèmes aléatoires (UMR CNRS 6074 - Université de Rennes 1 - INRIA - INSA de Rennes - ENS de Cachan - Télécom Bretagne - Université de Bretagne Sud)
Video services, Genetic algorithms, Network optimization