No abstract

Abstract-Streaming media on the internet has experienced rapid growth over the last few years and will continue to increase in importance as broadband technologies and authoring tools continue to improve. As the internet becomes an increasingly popular alternative to traditional communications media, internet streaming will become a significant component of many content providers' communications strategy. Internet streaming, however, poses significant challenges for content providers since it has significant distribution problems. Scalability, quality, reliability, and cost are all issues that have to be addressed in a successful streaming media offering.Streaming Content Delivery Networks attempt to provide solutions to the bottlenecks encountered by streaming applications on the internet. However only a small number of them has been deployed and little is known about the internal organization of these systems. In this paper we discuss the design choices made during the evolution of Akamai's Content Delivery Network for Streaming Media. In particular we look at the design choices made to ensure the network's scalability, quality of delivered content, and reliability while keeping costs low. Performance studies conducted on the evolving system indicate that our design scores highly on all of the above categories.

Abstract-

Abstract-The aim of this paper is to simulate the interaction of a large number of TCP controlled flows and UDP flows sharing many routers/links, from the knowledge of the network parameters (capacity, buffer size, topology, scheduling) and of the characteristics of each TCP (RTT, route etc.) and UDP flow. This work is based on the description via some fluid evolution equations, of the joint evolution of the window sizes of all flows over a single bottleneck router/link, as function of the synchronization rate. It is shown that the generalization of this fluid dynamics to a network composed of several routers can be described via equations allowing one to simulate the interaction of e.g. millions of TCP flows on networks composed of tens of thousands of links and routers on a standard workstation. The main output of the simulator are the mean value and the fluctuations of the throughput obtained by each flow, the localization of the bottleneck routers/links, the losses on each of them and the time evolution of aggregated input traffic at each router or link. The method is validated against NS simulations. We show that several important statistical properties of TCP traffic which were identified on traces are also present on traffic generated by our simulator: for instance, aggregated traffic generated by this representation exhibits the same short time scale statistical properties as those observed on real traces. Similarly, the experimental laws describing the fairness of the bandwidth sharing operated by TCP over a large network are also observed on the simulations.