Paper detail

Content-aware Traffic Engineering

Today, a large fraction of Internet traffic is originated by Content Providers (CPs) such as content distribution networks and hyper-giants. To cope with the increasing demand for content, CPs deploy massively distributed infrastructures. This poses new challenges for CPs as they have to dynamically map end-users to appropriate servers, without being fully aware of network conditions within an ISP as well as the end-users network locations. Furthermore, ISPs struggle to cope with rapid traffic shifts caused by the dynamic server selection process of CPs. In this paper, we argue that the challenges that CPs and ISPs face separately today can be turned into an opportunity. We show how they can jointly take advantage of the deployed distributed infrastructures to improve their operation and end-user performance. We propose Content-aware Traffic Engineering (CaTE), which dynamically adapts the traffic demand for content hosted on CPs by utilizing ISP network information and end-user location during the server selection process. As a result, CPs enhance their end-user to server mapping and improve end-user experience, thanks to the ability of network-informed server selection to circumvent network bottlenecks. In addition, ISPs gain the ability to partially influence the traffic demands in their networks. Our results with operational data show improvements in path length and delay between end-user and the assigned CP server, network wide traffic reduction of up to 15%, and a decrease in ISP link utilization of up to 40% when applying CaTE to traffic delivered by a small number of major CPs.