TCP Reno

TCP Reno is a TCP variant developed for congestion control and holds a significant place in modern computer networks. It emerged as a continuation of TCP Tahoe, which was created in the late 1980s to address congestion collapse in computer networks. TCP Reno employs the "Fast Retransmit" and "Fast Recovery" mechanisms to detect congestion during data transmission more quickly and improve transmission efficiency. This version was widely implemented in traditional wired networks and marked a major advancement in the evolution of TCP’s congestion control mechanisms.

The core operation of TCP Reno is based on interpreting packet loss as an indicator of network congestion. Reno continuously increases the transmission window (congestion window – cwnd); however, when a packet loss is detected, it reduces the window size. Two fundamental principles govern this process: additive increase and multiplicative decrease.

TCP Reno initiates a rapid retransmission process upon receiving three duplicate acknowledgments (DUPACKs) for the same segment. This mechanism allows the lost segment to be retransmitted without waiting for a timeout, significantly reducing the time required to detect and respond to congestion.

Unlike TCP Tahoe, TCP Reno does not revert directly to the Slow Start phase after Fast Retransmit. Instead, it activates the Fast Recovery algorithm, which follows these steps:

• Upon receiving three DUPACKs, the threshold value (ssthresh) is set to half the current congestion window size.
• The lost segment is retransmitted immediately.
• For each additional DUPACK received, the cwnd is increased by one segment size above ssthresh.
• When a new acknowledgment (ACK) arrives, the window size is set to ssthresh.

^{TCP Reno’s Congestion Window Behavior. (Drawn by YZ)}

Through this approach, TCP Reno reduces transmission speed in a controlled manner without fully dropping it, enabling faster recovery.

• Low Latency: Enables rapid recovery after packet loss, minimizing connection interruption time.
• Better Bandwidth Utilization: Allows the window to grow without returning to Slow Start, facilitating more efficient use of available bandwidth.

• Poor Fairness: Fails to ensure fair bandwidth sharing among connections with different delays. Connections with shorter delays tend to capture more bandwidth than those with longer delays.
• Over-Sensitivity to Packet Loss: Since packet loss is interpreted solely as a congestion signal, performance can degrade significantly in wireless networks or environments with high error rates.

Traditional TCP versions, including TCP Reno, encounter various performance issues in wireless networks. TCP Reno misinterprets packet loss occurring due to physical layer errors as congestion, leading to unnecessary interventions. In such cases, it unnecessarily reduces the window size, resulting in inefficient utilization of available bandwidth.