Internet QoS: a reality or daydreaming?

Telbisz Ferenc <telbisz@sunserv.kfki.hu>

KFKI RMKI SzHK és MATÁV PKI-FI

Internet was and is still working up to now as a "best effort" service, but more and more discussion is going on about the quality of service (QoS) of data communication on the Internet. This requirement is formulated by the customers, who need a reliable, high quality network for the their applications and the new applications, like video conferencing, VoIP, etc. put special requirements with respect to data transmission.

It was recognised naturally by IETF, too, that the accustomed quality parameters of Internet are not sufficient for the new applications. Thus QoS was dealt by two working groups investigating two different solutions. The first solution was the so called intserv (integrated service) method, where the communicating entities try to reserve the resources needed for the communication session. It turned out, however, that its scaling properties are rather bad, and thus it can not be applied with the backbone providers. The other investigated solution was the diffserv (differentiated services) method. The basic idea of this method is that the traffic is classified into different classes and the necessary resources are also allocated to these classes. The presentation will summarise shortly the main characteristics of these solution as well as it will give a critical analysis of them. As IP is a connectionless datagram service, IP QoS is always global (you can not speak of the quality of a given connection) and statistical, i.e. the quality parameters are fulfilled only on the average.

In what follows the presentation summarises the tools and methods available for providing differentiated services and their limitations. MPLS will be also discussed and its suitability for QoS services with the help of traffic engineering.

Probably this problem will be solved for a while by the aboundance of bandwidth available in the form of DWDM, as it will be higher at least by an order of magnitude as compared to the needs in the next future.