Output details

Transport Protocol Throughput Fairness

<05>This disruptive paper is the first to give analytical and empirical proof of why a widely-used performance metric, Jain's Fairness Index (JFI), can be inappropriate. We define a new metric, Normalised Resource Usage (NRU), and give an empirical comparison with JFI. Our evaluation includes mature and new protocols. We show more meaningful evaluation using NRU compared with JFI, and demonstrate possibilities for erroneous assessment with JFI. Our extensive empirical analyses of five popular transport protocols use configurations that compare real-world behaviour on a testbed. We also use emulations, where operational code is integrated into a widely-used simulation engine.