What triggers fast retransmit and fast recovery in TCP, and how do they work?

• A. Three duplicate ACKs trigger fast retransmit; fast recovery adjusts cwnd and ssthresh to recover quickly without full backoff after a loss
• B. A timeout triggers both fast retransmit and fast recovery, restarting the connection
• C. Two duplicate ACKs trigger fast retransmit and the others have no effect
• D. They are triggered by a connection close and do not apply during normal operation

Answer: (A)

Fast retransmit and fast recovery respond to packet loss without waiting for a timeout. When the sender observes three duplicate acknowledgments for the same outstanding segment, it infers that that specific segment was lost but subsequent data probably still arrives, so it immediately retransmits the missing segment. That quick retransmission is fast retransmit.

After performing the fast retransmit, the sender switches to fast recovery instead of dropping into a full backoff. It halves the current congestion window by updating the slow-start threshold to cwnd/2, then sets the congestion window to that new threshold plus a small additive amount to keep the pipeline busy (typically cwnd = ssthresh + 3*MSS). For each additional duplicate ACK, cwnd is increased by MSS to allow more retransmissions in flight. When a new ACK arrives that acknowledges the data beyond the retransmitted segment, the sender exits fast recovery and resumes normal congestion avoidance, setting cwnd to the updated ssthresh. In contrast, a timeout signals a more serious congestion event and leads to a full backoff (reducing cwnd to 1 MSS and entering slow start), rather than fast recovery.