delaybased

Delaybased refers to approaches and systems that base their control decisions on observed delays rather than solely on throughput or packet loss. In networking, delaybased methods monitor queueing delay or measured round‑trip time to infer current congestion and adjust transmission rate to keep delays low. They aim to reduce latency and improve responsiveness, particularly for delay‑sensitive applications, by preemptively reducing sending rate before packet loss occurs.

A classic example is TCP Vegas, a congestion control algorithm from the 1990s that compares the observed throughput to an estimated target and uses the estimated queuing delay to adjust the congestion window. More recent delaybased schemes include LEDBAT, which seeks to utilize spare network capacity while keeping queuing delays near a low, user‑defined target, signaling congestion primarily through delay rather than loss.

Delaybased concepts are also applied in routing and scheduling, where path or queueing delays serve as metrics to select routes or allocate resources. Across domains, delaybased control can yield lower latency and smoother performance when traffic mixes with loss‑based or aggressive flows, but it can be sensitive to measurement noise and delay variability.

Implementation challenges include accurate delay estimation, handling asymmetric paths, differentiating propagation delay from queueing delay, and adapting to rapidly changing networks. In practice, delaybased methods are often combined with other signals (loss, utility, or throughput) to balance responsiveness and fairness.

In summary, delaybased describes a family of approaches prioritizing delay measurements as the primary input for control decisions in networks and related systems, with the goal of reducing latency while maintaining efficient utilization.