ns Controlled Tests

The goal of our simulation is to validate our hypotheses in a controlled environment. In particular, we wish to i) test the non-interference property of Nice and ii) determine if Nice gets any useful bandwidth for the workloads considered. By using controlled ns [36] simulations in this phase of the study we can stress test the system by varying network configurations and load to extreme values. We can also systematically compare the Nice algorithm against others. Overall, the experiments support our theses:

• Nice flows cause almost no interference irrespective of the number of flows.
• Nice gets a significant fraction of the available spare bandwidth.
• Nice performs better than other existing protocols, including Reno, Vegas, and Vegas with reduced $\alpha$ and $\beta$ parameters.