Why TCP timers (still) don’t work well

We argue that the design principles of the TCP timeout algorithm are based solely on RTT estimations and may lead to flow synchronization, unnecessary retransmission effort and unfair resource allocation. We present a new Window-Based Retransmission Timeout algorithm (WB-RTO) for TCP, which exhibits two major properties: (i) it cancels retransmission synchronization, which dominates when resource demand exceeds resource supply and (ii) it reschedules flows on the basis of their contribution to congestion. WB-RTO achieves better fairness and slightly better goodput with significant less retransmission effort.