Evaluation for cost-effective cache deployment in ISP networks

Abstract

Internet traffic for video content delivery has grown rapidly because consumer-generated media has been shared throughout the world and the number of contents has increased tremendously. In response to this, ISPs have faced to the issue that how to deliver the massive contents with low cost and without losing network quality. The issue is significant for the ISPs providing Internet access service at a flat fee. One of the solutions to the problem is to cache contents on the servers located in ISP networks. The cache technology itself was developed from ten-odd years ago and several research works were done according to the traffic model in those days. However, most of works focused on the traffic between network providers; i.e. the traffic of inter-domain links. Although ISPs has the issue to reduce intra-domain links as well as inter-domain links, previous works have not answered the question that where cache servers should be located in ISP networks. In this paper, we analyzed the traffic in ISP networks with various conditions on the deployment of cache servers. The simulation considered the network topology including core and access networks, the disk capacity of cache servers and users' access model for requesting contents considering the unbalanced distribution of the populations on geographic locations. As one of results, we found that the traffic of the heaviest used link when the cache servers were deployed on the core network in city area was reduced 38.89% by the ratio of cache server disk size was 1 comparing to the case that the ratio of the cache server disk size was 4.