TCP Congestion Control

Summary

This video explains how TCP controls network congestion through a series of strategies, including Slow Start and Congestion Avoidance. The process starts with the congestion window growing exponentially during Slow Start, and once a threshold is reached, it switches to an additive increase in the Congestion Avoidance phase. If congestion is detected (when acknowledgments are not received), the window size is reduced by half, and the process restarts. Using the Additive Increase Multiplicative Decrease (AIMD) strategy, TCP efficiently manages network traffic to prevent congestion while ensuring optimal data flow.

Takeaways

• 😀 Congestion in computer networks happens when the number of packets exceeds the network's capacity, causing blockage.
• 😀 Congestion is typically caused by routers having finite buffer capacities to store packets.
• 😀 When a router's buffer is full, packets are dropped, which can lead to network congestion.
• 😀 TCP controls congestion by retransmitting segments if it doesn't receive an acknowledgment, further contributing to congestion.
• 😀 TCP detects congestion when it doesn't receive an acknowledgment for a packet within the estimated timeout (Round Trip Time, RTT).
• 😀 TCP uses the Additive Increase Multiplicative Decrease (AIMD) strategy to control congestion.
• 😀 The slow start algorithm initially sets the congestion window to 1 Maximum Segment Size (MSS), meaning the sender can send only one segment.
• 😀 As ACKs arrive, the congestion window increases exponentially, doubling each time until a threshold is reached.
• 😀 Once the congestion window reaches the threshold, the slow start phase stops, and the congestion avoidance phase begins.
• 😀 In the congestion avoidance phase, the congestion window increases additively by 1 MSS per round-trip time (RTT) instead of exponentially.
• 😀 If no acknowledgments are received within the timeout period, TCP assumes congestion has occurred, and the congestion window is reduced by half (multiplicative decrease), then it restarts the slow start phase.

Q & A

What is network congestion and how does it occur?

Network congestion occurs when the number of packets sent to the network exceeds its capacity. It typically happens because routers have a finite buffer to store packets. When more packets arrive than can be stored, packets are dropped, leading to congestion.

How does TCP detect congestion?

TCP detects congestion when it fails to receive an acknowledgment for a packet within the estimated Round Trip Time (RTT). This delayed acknowledgment indicates potential congestion in the network.

What is the Additive Increase Multiplicative Decrease (AIMD) strategy in TCP?

The AIMD strategy is used by TCP to control congestion. It increases the congestion window additively and decreases it multiplicatively when congestion is detected.

What is the Slow Start algorithm in TCP?

The Slow Start algorithm initially sets the congestion window to 1 Maximum Segment Size (MSS), meaning only one packet is sent. As ACKs are received, the congestion window increases exponentially, allowing more packets to be sent.

How does the congestion window grow during the Slow Start phase?

During the Slow Start phase, the congestion window grows exponentially. For example, if the window starts at 1 MSS, after receiving an acknowledgment, it increases to 2 MSS, then 4 MSS, and so on, doubling each time.

Why is there a threshold in the Slow Start phase?

The threshold is implemented to stop the exponential growth of the congestion window and avoid overwhelming the network. Once the congestion window reaches the threshold, TCP transitions to the Congestion Avoidance phase.

What happens when TCP transitions to the Congestion Avoidance phase?

In the Congestion Avoidance phase, the congestion window increases additively rather than exponentially. This means the window size increases by 1 MSS for each round of acknowledgments received, reducing the risk of congestion.

How does the congestion window behave in the Congestion Avoidance phase?

In the Congestion Avoidance phase, the congestion window increases linearly (additively), with each round of acknowledgments causing the window size to grow by 1 MSS until congestion is detected.

What occurs when congestion is detected in TCP?

When congestion is detected, the threshold size is reduced to half, and the congestion window is reset to 1 MSS. This triggers a return to the Slow Start phase, and the process repeats.

How does TCP control congestion through the combination of Slow Start and Congestion Avoidance?

TCP controls congestion by using the Slow Start algorithm for initial exponential growth and then switching to Congestion Avoidance to slow down the growth of the congestion window. If congestion is detected, TCP reduces the window and threshold, restarting the process to adapt to network conditions.