A Case Study-based Performance Evaluation Framework for CSCF Processes on a Blade-Server

The study of protocol behavior and traffic characteristics in a simulated environment is commonly supported by ad-hoc or general purpose simulators (e.g., Opnet, NS-2). But as the complexity of the network to be simulated grows, event-driven simulation could become the bottleneck of any performance analysis. Here, we propose an analytical technique for a real system performance evaluation of the proxy(P)-CSCF and serving(S)-CSCF components of an IMS network on a dual processor blade server. Based on this approach, we derive a specific model which can be applied to different network scenarios for performance evaluation and system engineering purposes. In particular, we study the optimal allocation of the traffic load in a blade-server generated by the CSCF processes to predict the maximum system capacity. The results indicate that our methodology (a) is effective in maximizing the blade utilization (b) points toward the benefits of hyper-threading and identifies the appropriate scheduling discipline at the local Linux schedulers and (c) provides a quantification of the optimal number of blades required to serve a predefined P/S-CSCF process arrival rate.