Stephan Jay, Thomas Plückebaum
Strategien zur Realisierung von Quality of Service in IP-Netzen
No. 315 / December 2008
Summary
IP-networks carry a multitude of different services on a single platform. These servicesexpress heterogeneous requirements regarding data rate and real-time transport. Accordinglyit needs to be determined how an appropriate Quality of Service (QoS) foreach of these services can be realized, what costs such an implementation incurs andwhat the incorporation of QoS implies for regulation and competition policy.
First, this study analyses the strengths and weaknesses of the three major QoSstrategiesover-engineering, reservation (Integrated Services & Resource ReserVationProtocol) and prioritisation (Differentiated Services). It is shown that QoS can be realizedby simply ensuring enough spare capacity within the network in order to avoid congestion. However, this is not always the most cost-effective approach, especially consideringthe limits to capacity expansion in last mile and (to a lesser extent) in aggregationnetworks. The advantages of differentiating QoS-architectures (reservation andprioritisation) are their ability to provide deterministic or statistic guarantees for QoSeven during High Load Periods. Reason to reservation’s weak scalability it is onlysparely implemented.
In a second step a model-based cost-comparison of the three strategies is conductedfor the link between the DSLAM and the next aggregation node. The calculation focuseson the bandwidth-driven cost and does not consider any potential add-on cost ofequipment and management that might be specific to the differentiating QoS strategies.
Analysis shows the dependency of scale effects and confirms the efficiency of the prioritisationstrategy that was already found in the qualitative analysis of the first step. Becauseof the scale effects in broadband networks and QoS provisioning regulatory authoritiesneed to take the cost of QoS into account when considering price decisions inorder to foreclose margin squeezing in QoS-enabled networks.In a third step the study analyzes pricing models for QoS services. There are a numberof reasons (danger of crowding out, welfare) that stand against applying Flat Rate pricingand support dynamic congestion pricing in QoS-enabled networks. However, analysis shows that there are significant barriers to the implementation of such pricing models. Accordingly the associated potentially problematic regulatory issues are not as relevantat present.
[only a German version available]
Discussion Paper is available for download.