Statistische Leistungsschranken für Computernetzwerke und Kommunikationssysteme

点击翻译按钮获取中文摘要

The realization of concepts with high scalability has been crucial to the success of the Internet. Consequently, backbone routers forward aggregated traffic, avoiding any per-flow treatment. Fairness among flows is approached by end-to-end congestion control, which, however, is not generally applicable. Hence, today's Internet implements only one basic service, called Best-Effort, that does not provide performance guarantees of any kind. The recent Differentiated Services architecture extends the single service Internet by adding few additional service classes, ideally with defined service guarantees, while maintaining the concept of traffic aggregation. However, in Aggregate scheduling networks the derivation of performance guarantees is complicated and still subject to research. The problem is caused by traffic aggregation and multiplexing effects that are hard to quantify. This project aims at evolving a framework for performance evaluation of Aggregate scheduling networks. The areas of application are manifold and range from dimensioning to admission control. A significant part of the intended work is the development of a concise probabilistic Network Calculus applying concepts from the theory of effective bandwidths. The goal is to extend today's worst-case analysis by allowing for certain violation probabilities.