Chapter 3: Transport Layer 3.5 - 3.7

Question 1

Characteristics of TCP connection oriented transport

Answer 1

• point to point (one sender one receiver)
• reliable, in order byte stream
• full duplex data
• pipelining, cumulative ACKs, and flow controlled

Question 2

TCP sequence number, ACKs ?

Answer 2

Sequence number : byte stream “number” of first byte in segment data

ACKs : sequence of next byte expected from the other side

Question 3

TCP round trip time ?

Answer 3

EstimatedRTT = (1 - a) * EstimatedRTT + (a * SampleRTT)
typical value of a = 0.125
SampleRTT : measured time from segment transmission until ACK received

Question 4

what is formula of TCP timeout interval

Answer 4

TimeoutInterval = EstimatedRTT + 4*DevRTT

where DevRTT …
DevRTT = (1-b)DevRTT + b |SampleRTT-EstimatedRTT|
typical value of b = 0.25

Question 5

TCP Sender events (simplified)

Answer 5

1. Data received from application
   
   • create segment with seq# and start timer if not
     already
   • seq# is a byte-stream number of first data byte in
     segment
2. timeout
   
   • retransmit segment that caused timeout
   • restart timer
3. ACK received
   
   • ifACK acknowledges previouslt unACKed segments:
     – update what is known to be ACKed
     =- start timer if there are still unACKed segments

Question 6

TCP fast retransmit

Answer 6

if sender receiver 3 ACKs with same data, resend unACKed segment with smallest seq#

Question 7

TCP flow control

Answer 7

receiver controls sender, so sender won’t overflow receiver’s buffer by transmitting too much too fast

Question 8

TCP 3 way handshake to open connection

Answer 8

1. Client sends SYN asking server to open a connection
2. Server receives SYN and send SYNACK saying ok so do you
3. Client responds with ACK saying yes, I’ll also open a connection then

Question 9

Congestion and how it is different from flow control

Answer 9

informally: too many sources sending too much data too fast for network to handle

differs from flow control, with fc it is one sender too fast for one receiver. With congestion, too many senders sending too fast.

Question 10

Some scenarios of causes of congestion

Answer 10

1. simple, one router and infinite buffers

2.

Question 11

congestion controls, why?

Answer 11

there are multiple flows in the internet, make sure every flow gets its fair share

Question 12

how TCP achieves congestion controls

Answer 12

making sure the sender increases transmission rate, probing for usual bandwidth until loss occurs

additive increase : increase cwnd(congestion window) by 1 MSS every RTT until loss detected

multiplicative decrease : cut cwnd in half after loss

Question 13

What cwnd (congestion window) contains ?

Answer 13

set of bytes that have been sent but not yet acknowledged

last byte sent - last byte acked <= cwnd

TCP sending rate : cwnd/RTT bps

Question 14

TCP slow start

Answer 14

Slow start ( a bit of misnomer):

• initial rate is slow but ramps up exponentially fast
• initially cwnd = 1MSS
• double cwnd every RTT (every ACK received)

Question 15

TCP detecting and reacting to loss

Answer 15

1. loss by timeout
   
   • cwnd set to 1 MSS
   • windows then grow exponentially (slow start)
2. loss indicated by 3 duplicate ACKs : TCP Reno
   
   • cwnd is cut in half and grow linearly
3. TCP Tahoe always sets cwnd to 1 (timeout or 3 duplicate ACKs)

Question 16

TCP Tahoe Congestion Control flow

Answer 16

1. Slow start until reaches threshold
2. Once reaches threshold, AIMD (additive increase, multiplicative decrease) kicked in
3. keeps increasing window size by 1 increment(congestion avoidance) .Until 1 of the segment is loss (by timeout or etc, say this is loss at segment 12 ) , then cwnd will be reset to 1.
4. slow start phase again , but now the threshold is half of when loss occurs (so 12/2 =6), after reaches 6 segments AIMD kicked in

Question 17

TCP Reno Congestion Control flow, (fast recovery)

Answer 17

similar with Tahoe, but when loss occurs by duplicate, instead of going to 1 immediately, it just goes half and increase linearly (also known as fast recovery)