Comparison of Marking Algorithms for PCN-Based Admission Control

The PCN working group of the IETF discusses the use of pre-congestion notification (PCN) to implement flow admission control. Packet meters and markers are used on all network links and packet markings are recorded as congestion level estimates (CLEs) the egress nodes. The working group currently discusses the pros and cons of possible marking algorithms that play a major role in this new architecture. This paper provides a detailed description of threshold and ramp marking based on a virtual queue formulation. We investigate the impact of the marking threshold and the virtual queue size on the marking behavior and develop different marking strategies. We test the robustness of the CLEs obtained for both marking schemes against different CLE parameters and traffic characteristics. Furthermore, we show that ramp marking can be well approximated by appropriately configured threshold marking.