In TCP congestion control, how does the congestion window (cwnd) grow during the 'Slow Start' phase?
During Slow Start, cwnd starts at 1 MSS and increments by 1 MSS for each received ACK, resulting in exponential doubling of the congestion window per Round-Trip Time (RTT).
What event triggers the transition of TCP congestion control from the 'Slow Start' phase to the 'Congestion Avoidance' phase?
When cwnd >= ssthresh, TCP switches from exponential growth (Slow Start) to linear growth (Congestion Avoidance).
During the 'Congestion Avoidance' phase of TCP Tahoe and Reno, how is the congestion window (cwnd) adjusted per RTT?
Congestion Avoidance employs Additive Increase: upon receiving a full window of ACKs, cwnd increases linearly by 1 Maximum Segment Size (MSS) per RTT.
In TCP Reno, what reaction takes place upon receiving 3 duplicate ACKs (Fast Retransmit / Fast Recovery)?
TCP Reno avoids dropping cwnd to 1 MSS on 3 duplicate ACKs; it sets ssthresh = cwnd / 2, cuts cwnd to ssthresh + 3 MSS, and continues with linear growth (Fast Recovery).
Why does the active closer in a TCP connection enter the TIME_WAIT state for 2MSL (Maximum Segment Lifetime)?
The TIME_WAIT state lasts for 2 * MSL to guarantee that the final ACK was delivered (retransmitting if the peer resends FIN) and to prevent old duplicate packets from interfering with a new incarnation of the connection.