Delay-based Congestion Control

This approach aims to keep the bottleneck link busy, but not full, thus avoiding high delay times and buffering.

$RTT_{min}=\textrm{The minimum observed RTT (uncongested path).}$

$\begin{indisplay}\begin{split} \textrm{Uncongested throughput }\lambda_+ &= \fra... ...frac{\textrm{\char93 of bytes sent}}{RTT_{measured}} \end{split}\end{indisplay}$

(6)

If $\lambda_{act}$ is close to $\lambda_+$ , then $cwnd+\beta$ , otherwise if $\lambda_{act}$ is far below $\lambda_+$ , then $cwnd-\beta$ . Where $\beta$ is some predefined constant.

This method controls congestion without forcing / inducing a loss event, and maximises throughput with minimal delays.