Rules for Faster Convergence to Fairness
Ageliki Tsioliaridou, Chi Zhang, Vassilis Tsaoussidis · 2009
We extend the theoretical model of Chiu and Jain described in [2] by taking into account the role of bottleneck bufier and present a new analysis that allows system to reach equilibrium fast, based on throughput measurements. In our model, no ∞ow is aware of the number of competitors in the channel, no ∞ow is aware of the bandwidth and bufier capacity; however, at the end of each epoch, which is signaled by a congestion event, each ∞ow can i) estimate whether it has operated above, beyond or around its fair-share and ii) adjust its window to its fair-share. Within this context, we present two rules on how to adjust the transmission rate after a congestion event, namely Fair-Share and Fairness rule. They both promote fairness without damaging system e‐ciency and responsiveness. Simulation results conflrm the validity of our analysis and the proposed congestion control schemes.