Reproducible Research in Network Modeling
This study addresses the longstanding limitation in networking research caused by the scarcity of reliable and reproducible experimental data. To overcome this challenge, the authors propose employing high-fidelity software models as substitutes for physical devices, enabling the construction of reproducible validation environments through “natural experiments” and establishing a reproducibility-based criterion for experimental reliability. A systematic evaluation framework is developed to conduct full-scale assessments of mainstream network simulation tools. The findings reveal that while most tools technically satisfy reproducibility requirements, their adoption in practice is predominantly influenced by non-technical factors such as popularity and user familiarity. This work introduces a new paradigm and benchmark for reproducible experimentation in networking research.