Master Class : TCP/IP Mechanics from Scratch to Expert

1. TCP Retransmission: TCP (Transmission Control Protocol) retransmits segments that are lost or corrupted during transmission. This is done to ensure data integrity and reliability in a communication channel that may have errors.

2. TCP RTO (Re Transmission TimeOut) Problems if computed Wrongly: The RTO is the time TCP waits before retransmitting a segment when no acknowledgment has been received. If the RTO is set incorrectly, it can lead to either excessive retransmissions (if too low) or inefficient use of bandwidth (if too high).

3. Expectations from TCP when Segment loss occurs: When a segment loss occurs, TCP expects that the segment is lost and needs to be retransmitted. It then waits for a certain amount of time (RTO) before retransmitting the segment. After retransmission, it awaits an acknowledgment for that segment.

4. TCP Exponential Backoff - When consecutive segment loss occurs: TCP uses an exponential backoff strategy when multiple consecutive segment losses are detected, which helps in avoiding the congestion collapse. The backoff algorithm doubles the RTO until an acknowledgment is received or the maximum RTO (max_rto) is reached.

5. TCP RTO Value Estimation: The initial RTO is estimated based on the round-trip time (RTT) of the last four ACKs, plus four times the standard deviation of those RTTs. The actual RTO is adaptively changed over time as more RTT measurements are gathered.

6. TCP Retransmission Ambiguity Problem: After a timeout event, there can be ambiguity in determining which segment to retransmit because multiple segments might have been lost during the same period. This is resolved by using sequence numbers and acknowledgments.

7. Karn's Algorithm: Karn's algorithm resolves the ambiguity of knowing which packet to retransmit after a timeout by retransmitting all packets that were unacknowledged when the timeout occurred, up to those that are no more than one RTT old.

8. Karn's Algorithm Illustration: The algorithm takes the clock (the last known delivered byte) and the sender's window as of the last ACK received, and retransmits all unacknowledged packets up to and including the first packet that is no more than one RTT old.

9. Concept of Fast Retransmission: Introduced in TCP Tahoe, fast retransmission is used when there are signs of persistent congestion or multiple duplicate acknowledgments (dups). It sends a segment that is believed to be missing more quickly than the regular timer-based approach would allow.

10. TCP handling out of order segments: TCP can handle out-of-order segments by keeping a buffer for them and reordering them before delivering them to the application layer.

11. TCP holes Problem and its remedy: A hole is a range of byte numbers where acknowledgments have been received but not all of the data for that range has been received. TCP fills these holes using selective acknowledgments (SACKs) or by retransmitting data within the hole during fast retransmit.

12. Redundant Retransmission due to dupACK: Duplicate ACKs (dupacks) are used to inform the sender that some of the segments have been received more than once. This can prevent unnecessary retransmissions if the sender were to naively interpret each ACK as indicating a different segment was missing.

13. TCP Congestion Control Algorithms: TCP has several congestion control algorithms:

• slow start
• congestion avoidance
• fast recovery
• fast retransmit

14. Typical TCP Graph: A typical TCP graph shows the cumulative number of bytes sent on one axis and time on the other. It often exhibits characteristics such as slow start, congestion avoidance, and periods after timeout events where fast recovery is used.

15. Algorithm Selection Flowchart: The flowchart for selecting the appropriate TCP congestion control algorithm includes conditions based on recent network performance (e.g., RTT, packet loss), such as whether to switch from slow start to congestion avoidance, or when to initiate fast retransmit or fast recovery.

TCP's robust design ensures that it can handle a variety of network conditions and recover gracefully from packet loss, out-of-order segments, and other issues that can arise during data transmission. Understanding these concepts is crucial for anyone working with networks or developing applications that require reliable data transmission over the internet. Good luck with your course!