In TCP congestion control, how do slow-start and congestion avoidance differ?

• A. Slow-start uses exponential growth; congestion avoidance uses linear growth after threshold.
• B. Slow-start uses linear growth; congestion avoidance uses exponential.
• C. Slow-start reduces cwnd on loss; congestion avoidance increases cwnd linearly.
• D. Slow-start increases cwnd exponentially until the slow-start threshold; congestion avoidance increases cwnd linearly after the threshold.

Answer: (D)

TCP congestion control probes available bandwidth with two distinct phases: slow-start and congestion avoidance. In slow-start, the congestion window (cwnd) starts small and grows rapidly each round-trip, effectively doubling as ACKs arrive. This exponential growth continues until cwnd reaches the slow-start threshold (ssthresh). At that point, the algorithm switches to congestion avoidance, where growth becomes linear, typically increasing cwnd by about one MSS per RTT. This slower, additive growth helps avoid overshooting the network capacity while still probing for more bandwidth.

So the best description is that slow-start increases cwnd exponentially until the slow-start threshold, and congestion avoidance increases cwnd linearly after the threshold. The other statements mix up the growth patterns or omit the threshold-driven transition, which is why they don’t fit.