Statistically Sound Experiments with OpenAirInterface Cloud-RAN Prototypes

Research on 4G/5G cellular networks is progressively shifting to paradigms that involve virtualization and cloud computing. Within this context, prototyping assumes a growing importance as a performance evaluation method, besides large-scale simulations, as it allows one to evaluate the computational requirements of the system. Both approaches share the need for a structured and statistically sound experiment management, with the goal of reducing errors in both planning and measurement collection. In this paper, we describe how we solve the problem with OpenAirInterface (OAI), an open-source system for prototyping 4/5G cellular networks. We show how to integrate a sound, validated software, namely ns2-measure, with OAI, so as to enable harvesting samples of arbitrary metrics in a structured way, and we describe scripts that allow structured experiment management, such as launching a parametric simulation campaign and harvesting its results in a plot-ready format. We complete the paper by demonstrating some advantages brought about by our modifications.