Pre-Congestion Notification (PCN) has been recently proposed at IETF as a practical framework for scalable measurement-based flow control system for support of inelastic real-time traffic. PCN encompasses two complementary functions: admission control of new flows and termination of some admitted flows when the available capacity is unexpectedly reduced after a network failure. Its simple structure and ad-hoc measurementbased nature warrant careful exploration of its applicability. In this work, we conducted a comprehensive performance evaluation of two proposals that implement PCN, CL-PHB and Single-Marking. We showed that CL-PHB works reliably for both admission and termination control, which in turn demonstrates that the PCN framework is a viable architecture to provide QoS to inelastic real-time traffic. We also studied Single-Marking's performance tradeoffs associated with its implementation benefits, and analyzed its applicability range. Our work also includes the development of several novel fluid models of different complexities and the investigation of their accuracy and applicability to various aspects of PCN admission and termination control. Aside from providing a tool to gain more insight in PCN behavior, we believe that some of our analysis is of a more general nature. Specifically, we believe that our results support a conjecture that while bottleneck behaviors are possible to predict with high accuracy using these models, per-aggregate behavior of admission control systems cannot be accurately predicted with any fluid model resulting in a bounded estimation error.