Evolution of sender’s congestion window in Window-Based End-to-End Congestion Control

[HelloIN]

We have the attached network diagram

 

The mobile browser (TCP receiver) wants to download a webpage from a server hosting example.com

 

Packet loss can only happen when the queue on the Wi-Fi router overflows (> 100 packets). The link speeds are in Mega bits per second, where 1 Mbps = 1,000,000 bits per second. Also, assume that the Wi-Fi router adds no hop delay or queuing delay (no matter the queue size). We consider the size of each data packet to be 1,500 bytes. Furthermore, assume that each packet can carry 1,500 bytes of TCP payload. For example, the sender would need 2 packets to transmit a webpage of 3 KB (3,000 bytes) to the receiver.

 

The sender uses the following slow start mechanism which proceeds in time intervals of the RTT: (i) In the first RTT interval, the sender’s initial cwnd is equal to 2 packets. The sender bursts these 2 packets and receives corresponding 2 ACKs from the receiver. (ii) For each subsequent RTT interval, the sender increases the cwnd by 1 packet for every ACK it receives during the previous RTT interval. (iii) The receiver always sends 1 ACK for each packet received from the sender. This means that, the sender would effectively double its cwnd in each subsequent RTT interval. (iv) The sender would exit the slow start and switch to congestion avoidance if/when it detects the first packet loss.

 

 

Can someone please help me understand in which RTT will the queue in the WiFi router start building up (queue size greater than 0)? In RTT #1, the sender sends 2 packets and receives 2 ACKs.

 

Further, In which RTT will the first packet drop occur at the WiFi router?