What is the retransmission timeout (RTO) and how is it calculated?

• A. A constant fixed timer chosen arbitrarily by the sender
• B. A timer derived from the average RTT and its variance using adaptive algorithms
• C. The timer value for retransmission; calculated from RTT estimates using algorithms that track average RTT and its variance (e.g., Jacobson/Karn)
• D. A timer that only applies when keep-alives are enabled

Answer: (C)

RTO is the timer that determines when a TCP sender should retransmit a segment after not receiving its acknowledgment. Because network RTT changes over time, you don’t use a fixed, arbitrary value. Instead, you continuously estimate the round-trip time and its variability, and you compute the timeout from those estimates. In TCP this is done with RTT samples to form a smoothed RTT (SRTT) and a measure of variation (RTTVAR). The retransmission timeout is then set from those values, typically using a formula like RTO = SRTT + 4 × RTTVAR (with a small clock-granularity term in some implementations). A key detail from the classic approach is to avoid using RTT samples from a retransmission itself (Karn’s rule), to prevent biasing the estimate.

This adaptive approach allows the timer to respond to changing conditions: shortens when the path is fast and stable, and increases when delays become bursty, reducing unnecessary retransmissions. The RTO is specifically for retransmission timing and is not tied to keep-alives, which are separate mechanisms.