REPRODUCIBLE NETWORK RESEARCH WITH A HIGH-FIDELITY SOFTWARE-DEFINED NETWORK TESTBED

Network research, as an experimental science, ought to be reproducible. However, it is not a standard practice to share models, methods or software code to support experimental evaluation and reproducibility when publishing a network research paper. In this work, we advocate reproducible networking experiments by building a unique testbed consisting of container-based network emulation and physical devices. The testbed provides a realistic and scalable platform for reproducing network research. The testbed supports large-scale network experiments using lightweight virtualization techniques and capable of running across distributed physical machines. We utilize the testbed to reproduce network experiments, and demonstrate the e↵ectiveness by comparing the results with the original published network experiments, such as Hedera, a scalable and adaptive traffic flow scheduling system in data center networks.M.S. in Computer Science, May 201