Quality of Service in the Internet Age

Abstract – Quality of Service (QoS) is basically the attempt to guarantee performance metrics such as packet delay, jitter, packet loss or throughput under worst-case conditions on a network. This is achieved by carefully matching or reserving link bandwidth to the QoS requirements at hand. Logically, this can be achieved in connection-orientated networks only. Yet wide area IP networks are neither connection-orientated nor reserve bandwidth. The push to convergence of disparate technologies over a single network however means that workarounds must be found to practically enable QoS on IP networks. This paper discusses several mechanisms and protocols that work together to provide a realizable end-to-end QoS solution. Starting with classifiers, queues and polciers at switches one shall see how packets are scheduled to provide relative priority. A mathematical insight into benefits of inter-class resource sharing is also presented, along with a discussion of where current theory falls short of operator requirements. Lastly various protocols such as IntervServ, RSVP, DiffServ, MPLS and traffic engineering extensions are introduced, and the manner in which they fit together illustrated