Capacity management for Internet traffic

We describe methods to guarantee a certain level of service for Internet traffic by reserving capacity along fixed logical paths. The amount of needed capacity is calculated by using a trade-off between connection rejection probability and the utilization of the capacity. This can be used in the process of dimensioning capacity on a long term basis. Two dynamic allocation methods are proposed, which periodically reallocates capacity according to measured traffic loads. Call admission control is used and automatic connection retrials are studied. Each method is designed for a particular scenario. In the first one, a LAN reserves capacity to get a certain transmission speed for every connection. Capacity is paid for according to how much of it that is reserved. The second one considers an aggregated traffic generated by users having limited transmission speed in their connection to the Internet (e.g. modern- and mobile-users) and the operator of the network manages the capacity. The Internet traffic is modeled as Web page fetches from the World Wide Web.