Robust routing in ISP networks considering YouTube traffic demand fluctuations

Content Delivery Networks (CDN) are responsible for a large traffic share in today´s ISP networks. CDN operators deploy load balancing between their data centers which may cause large traffic fluctuations within an ISP network. These traffic shifts in turn can lead to severe congestion situations if the traffic routing is not able to distribute the load properly. In this paper we first present a traffic demand model that includes the representation of YouTube traffic fluctuations caused by the dynamic data center selection mechanism applied in the YouTube CDN. This model is derived from measurements of the YouTube data center selection behavior and from traffic measurements in a large German ISP network. We then address the problem of determining a robust routing scheme using our traffic demand model. The target of the routing optimization is to minimize the maximum link utilization in an ISP network with given link capacities. Finally we apply our robust routing approach on different network topology and traffic demand scenarios and discuss the results.