Transport Layer Protocol Practice Questions

The Mathis formula for TCP throughput: Throughput ≈ (1.22 × MSS) / (RTT × √p) where p is the packet loss probability and RTT is the round-trip time. This formula shows: (1) Throughput is inversely proportional to RTT - slower links (higher RTT) have lower throughput. (2) Throughput is inversely proportional to √p - higher loss rates significantly reduce throughput. (3) To double throughput, loss rate must be reduced by a factor of 4. This formula captures the AIMD sawtooth behavior averaged over many congestion cycles.

For Go-Back-N with $n$-bit sequence numbers, the maximum sender window size is $2^n - 1$. With $n = 4$, the maximum window size is $2^4 - 1 = 16 - 1 = 15.

UDP (User Datagram Protocol) is connectionless and unreliable, offering minimal overhead with no connection setup, no acknowledgment, and no retransmission. It is ideal for applications like DNS (fast request-response) and real-time streaming where low latency matters more than guaranteed delivery.

The TCP sender limits its outstanding unacknowledged data to the minimum of the receiver's advertised window (rwnd) and the congestion window (cwnd): effective window = min(rwnd, cwnd). rwnd reflects the receiver's buffer availability (flow control). cwnd reflects the sender's estimate of the network's capacity (congestion control). Using the minimum ensures neither the receiver nor the network is overwhelmed. The sending rate ≈ min(rwnd, cwnd) / RTT.

Token bucket allows bursts up to the bucket size (tokens accumulate when idle), while the leaky bucket enforces a constant output rate and drops excess packets, not allowing any bursts.

In Slow Start, cwnd doubles every RTT: RTT 1: $cwnd=1$, RTT 2: $cwnd=2$, RTT 3: $cwnd=4$, RTT 4: $cwnd=8$, RTT 5: $cwnd=16$. At $cwnd=16$, which equals ssthresh, TCP transitions to Congestion Avoidance. Therefore, $5$ RTTs are needed.

TCP header contains Sequence Number, Acknowledgment Number, and Window Size fields for reliable delivery and flow control. UDP header is minimal, containing only Source Port, Destination Port, Length, and Checksum. Therefore, Sequence Number, Acknowledgment Number, and Window Size are TCP-only fields.

TCP ensures reliable, ordered delivery.

Slow Start (cwnd doubles each RTT until cwnd ≥ ssthresh = 64): RTT1:2, RTT2:4, RTT3:8, RTT4:16, RTT5:32, RTT6:64 (reaches ssthresh). Switch to Congestion Avoidance (+1 MSS/RTT): RTT7:65, RTT8:66, RTT9:67, RTT10:68, RTT11:69, RTT12:70. Wait - at RTT6 cwnd=64 which equals ssthresh, so CA starts at RTT7. RTT7:65, RTT8:66, RTT9:67, RTT10:68, RTT11:69, RTT12:70. cwnd reaches 70 at RTT12. But counting from RTT1: 6 slow start RTTs + 6 CA RTTs = 12 RTTs. Actually at RTT6 cwnd=64 (Slow Start: 1→2→4→8→16→32→64, that's 6 RTTs). Then CA: RTT7:65,...,RTT12:70. Total = 12 RTTs to reach 70. Hmm, need to reach first ≥ 70: at RTT 12, cwnd = 70. Answer = 12.

In the 3-way handshake: (1) Client sends $SYN=1, ACK=0$; (2) Server replies with $SYN=1, ACK=1$; (3) Client sends $ACK=1, SYN=0$. Simultaneous SYN and FIN is not a standard valid state in TCP connection management. Option D describes an invalid combination for normal operation.

A: TRUE - standard TCP (Reno, CUBIC) infers congestion from packet loss (timeout or duplicate ACKs) since routers do not explicitly notify senders. B: FALSE - TCP Reno treats timeout as more severe (cwnd→1 MSS, Slow Start restart) than 3 dup ACKs (cwnd halved, Fast Recovery - stays in Congestion Avoidance). C: TRUE - TCP CUBIC replaces the linear increase of Reno with a cubic function, allowing faster recovery of bandwidth in high BDP (bandwidth-delay product) networks like long-fat pipes. D: TRUE - TCP Vegas monitors RTT inflation (increasing RTT signals growing queues) to reduce cwnd before packet loss, achieving smoother control and avoiding the loss-based sawtooth pattern.

On timeout in TCP Reno: ssthresh = max(cwnd/2, 2 MSS) = max(16/2, 2) = max(8, 2) = 8 MSS. cwnd = 1 MSS (reset to Slow Start). The connection re-enters Slow Start from cwnd = 1 MSS and will grow exponentially until cwnd reaches the new ssthresh = 8 MSS, then switch to Congestion Avoidance. New values: cwnd = 1 MSS, ssthresh = 8 MSS.

Selective Repeat individually acknowledges all correctly received packets and the sender only retransmits lost/damaged packets. To avoid sequence number ambiguity, both sender and receiver windows must be at most $2^{n-1}$. Unlike Go-Back-N, SR buffers out-of-order packets at the receiver rather than discarding them.

Stop-and-Wait ARQ: the sender transmits exactly one frame, then waits for an acknowledgment (ACK) before transmitting the next frame. If the ACK is not received within the timeout period, the frame is retransmitted. This is the simplest ARQ protocol but is extremely inefficient for high bandwidth-delay product links because the sender is idle for 2×propagation delay after each frame. It is equivalent to a sliding window protocol with window size W = 1. Utilization = T/(T+2D) = 1/(1+2a).

UDP checksum is computed over the UDP header, the UDP payload data, and a pseudo-header that contains fields from the IP header (source IP, destination IP, protocol number, and UDP length). This provides additional integrity verification.

TCP Reno distinguishes two loss signals: Timeout (severe congestion): ssthresh = cwnd/2, cwnd = 1 MSS, enter Slow Start. This aggressive response reflects that a timeout indicates severe congestion (all in-flight packets may be lost). Three duplicate ACKs (moderate congestion, Fast Retransmit): ssthresh = cwnd/2, cwnd = ssthresh (= cwnd/2), enter Congestion Avoidance directly - this is Fast Recovery. Three dup-ACKs indicate one packet was lost but later packets are arriving (network still functional), so a less drastic reduction is appropriate.

Slow Start: cwnd is initialized to 1 MSS (Maximum Segment Size). For each ACK received, cwnd increases by 1 MSS - since each RTT produces as many ACKs as segments sent, cwnd effectively doubles per RTT (exponential growth). Slow Start continues until: (1) cwnd reaches ssthresh (slow start threshold) → switches to Congestion Avoidance (linear increase), or (2) a timeout occurs → cwnd reset to 1 MSS and ssthresh halved, or (3) three duplicate ACKs → Fast Retransmit/Recovery. Despite the name 'slow', cwnd grows exponentially - 'slow' refers to starting from 1 MSS rather than the full window.

UDP is connectionless.