Proportional rate reduction for TCP

Packet losses increase latency for Web users. Fast recovery is a key mechanism for TCP to recover from packet losses. In this paper, we explore some of the weaknesses of the stan-dard algorithm described in RFC3517 and the non-standard algorithms implemented in Linux. We find that these algo-rithms deviate from their intended behavior in the real world due to the combined effect of short flows, application stalls, burst losses, acknowledgment (ACK) loss and reordering, and stretch ACKs. Linux suffers from excessive congestion window reductions while RFC3517 transmits large bursts under high losses, both of which harm the rest of the flow and increase Web latency. Our primary contribution is a new design to control trans-mission in fast recovery called proportional rate reduction (PRR). PRR recovers from losses quickly, smoothly and accurately by pacing out retransmissions across received ACKs. In addition to PRR, we evaluate the TCP early retransmit (ER) algorithm which lowers the duplicate ac-knowledgment threshold for short transfers, and show that delaying early retransmissions for a short interval is effec-tive in avoiding spurious retransmissions in the presence of a small degree of reordering. PRR and ER reduce the TCP latency of connections experiencing losses by 3-10 % depend-ing on the response size. Based on our instrumentation o