1

TCP - Part II

Relates to Lab 5. This is an extended module that covers TCP data transport, and flow control, congestion control, and error control in TCP.

2

Interactive and bulk data

TCP applications can be put into the following categoriesbulk data transfer - ftp, mail, httpinteractive data transfer - telnet, rlogin

TCP has algorithms to deal which each type of applications efficiently.

3

tcpdump of an rlogin session

This is the output of typing 3 (three) characters :44.062449 argon.cs.virginia.edu.1023 > neon.cs.virginia.edu.login: P 0:1(1) ack 1 44.063317 neon.cs.virginia.edu.login > argon.cs.virginia.edu.1023: P 1:2(1) ack 1 win 876044.182705 argon.cs.virginia.edu.1023 > neon.cs.virginia.edu.login: . ack 2 win 17520

48.946471 argon.cs.virginia.edu.1023 > neon.cs.virginia.edu.login: P 1:2(1) ack 2 win 1752048.947326 neon.cs.virginia.edu.login > argon.cs.virginia.edu.1023: P 2:3(1) ack 2 win 876048.982786 argon.cs.virginia.edu.1023 > neon.cs.virginia.edu.login: . ack 3 win 17520

55:00.116581 argon.cs.virginia.edu.1023 > neon.cs.virginia.edu.login: P 2:3(1) ack 3 win 1752055:00.117497 neon.cs.virginia.edu.login > argon.cs.virginia.edu.1023: P 3:4(1) ack 3 win 876055:00.183694 argon.cs.virginia.edu.1023 > neon.cs.virginia.edu.login: . ack 4 win 17520

Argon.cs.virginia.edu Neon.cs.virginia.edu

rlogin sessionfrom Argonto Neon

4

Rlogin

• “Rlogin” is a remote terminal application• Originally built only for Unix systems.• Rlogin sends one segment per character (keystroke)• Receiver echoes the character back.

• So, we really expect to have four segments per keystroke

5

Rlogin

• We would expect that tcpdump shows this pattern:

• However, tcpdump shows this pattern:

• So, TCP has delayed the transmission of an ACK

character

ACK of character

ACK of echoed character

echo of character

character

ACK and echo of character

ACK of echoed character

6

Delayed Acknowledgement

• TCP delays transmission of ACKs for up to 200ms• The hope is to have data ready in that time frame. Then, the

ACK can be piggybacked with the data segment.• Delayed ACKs explain why the ACK and the “echo of

character” are sent in the same segment.

7

tcpdump of a wide-area rlogin session

This is the output of typing 9 characters :

54:16.401963 argon.cs.virginia.edu.1023 > tenet.CS.Berkeley.EDU.login: P 1:2(1) ack 2 win 1638454:16.481929 tenet.CS.Berkeley.EDU.login > argon.cs.virginia.edu.1023: P 2:3(1) ack 2 win 1638454:16.482154 argon.cs.virginia.edu.1023 > tenet.CS.Berkeley.EDU.login: P 2:3(1) ack 3 win 1638354:16.559447 tenet.CS.Berkeley.EDU.login > argon.cs.virginia.edu.1023: P 3:4(1) ack 3 win 1638454:16.559684 argon.cs.virginia.edu.1023 > tenet.CS.Berkeley.EDU.login: P 3:4(1) ack 4 win 1638354:16.640508 tenet.CS.Berkeley.EDU.login > argon.cs.virginia.edu.1023: P 4:5(1) ack 4 win 1638454:16.640761 argon.cs.virginia.edu.1023 > tenet.CS.Berkeley.EDU.login: P 4:8(4) ack 5 win 1638354:16.728402 tenet.CS.Berkeley.EDU.login > argon.cs.virginia.edu.1023: P 5:9(4) ack 8 win 16384

argon.cs.virginia.edu tenet.cs.berkeley.edu

rlogin sessionbetween argon.cs.virginia.eduandtenet.cs.berkeley.edu

8

Wide-area Rlogin: Observation 1

• Transmission of segments follows a different pattern.

• The delayed acknowled-gment does not kick in

• Reason is that there is always data at aida when the ACK arrives.

char1

ACK of char 1 + echo of char1

ACK + char2

ACK + echo of char2

9

Wide-area Rlogin: Observation 2

• There are fewer transmissions than there are characters.• Aida never has multiple segments outstanding.• This is due to Nagle’s Algorithm:

Each TCP connection can have only one small (1-byte) segment outstanding that has not been acknowledged.

• Implementation: Send one byte and buffer all subsequent bytes until acknowledgement is received.Then send all buffered bytes in a single segment. (Only enforced if data is arriving from application one byte at a time)

• Nagle’s rule reduces the amount of small segments.The algorithm can be disabled.

10

Flow Control

Congestion ControlError Control

TCP:

11

What is Flow/Congestion/Error Control ?

• Flow Control: Algorithms to prevent that the sender overruns the receiver with information?

• Congestion Control: Algorithms to prevent that the sender overloads the network

• Error Control: Algorithms to recover or conceal the effects from packet losses

The goal of each of the control mechanisms is different.

But the implementation is combined

12

TCP Flow Control

13

TCP Flow Control

• TCP implements sliding window flow control

• Sending acknowledgements is separated from setting the window size at sender.

• Acknowledgements do not automatically increase the window size

• Acknowledgements are cumulative

14

Sliding Window Flow Control

1 2 3 4 5 6 7 8 9 10 11

Advertised window

sent but notacknowledged can be sent

USABLEWINDOW

sent andacknowledged

can't sent

• Sliding Window Protocol is performed at the byte level:

•Here: Sender can transmit sequence numbers 6,7,8.

15

Sliding Window: “Window Closes”

1 2 3 4 5 6 7 8 9 10 11

1 2 3 4 5 6 7 8 9 10 11

Transmit Byte 6

1 2 3 4 5 6 7 8 9 10 11

AckNo = 5, Win = 4is received

• Transmission of a single byte (with SeqNo = 6) and acknowledgement is received (AckNo = 5, Win=4):

16

Sliding Window: “Window Opens”

1 2 3 4 5 6 7 8 9 10 11

1 2 3 4 5 6 7 8 9 10 11

AckNo = 5, Win = 6is received

• Acknowledgement is received that enlarges the window to the right (AckNo = 5, Win=6):

• A receiver opens a window when TCP buffer empties (meaning that data is delivered to the application).

17

Sliding Window: “Window Shrinks”

1 2 3 4 5 6 7 8 9 10 11

1 2 3 4 5 6 7 8 9 10 11

AckNo = 5, Win = 3is received

• Acknowledgement is received that reduces the window from the right (AckNo = 5, Win=3):

• Shrinking a window should not be used

18

Window Management in TCP

• The receiver is returning two parameters to the sender

• The interpretation is:• I am ready to receive new data with

SeqNo= AckNo, AckNo+1, …., AckNo+Win-1

• Receiver can acknowledge data without opening the window• Receiver can change the window size without acknowledging

data

AckNowindow size

(win)32 bits 16 bits

19

Sliding Window: Example

3K

ReceiverBuffer

0 4KSendersends 2Kof data

2K

Sendersends 2Kof data

4K

Sender blocked

20

TCP Congestion Control

21

TCP Congestion Control

• TCP has a mechanism for congestion control. The mechanism is implemented at the sender

• The window size at the sender is set as follows:

where • flow control window is advertised by the receiver• congestion window is adjusted based on feedback from the

network

•Send Window = MIN (flow control window, congestion window)•Send Window = MIN (flow control window, congestion window)

22

TCP Congestion Control

• The sender has two additional parameters:– Congestion Window (cwnd)

Initial value is 1 MSS (=maximum segment size) counted as bytes

– Slow-start threshold Value (ssthresh)Initial value is the advertised window size)

• Congestion control works in two modes:– slow start (cwnd < ssthresh)– congestion avoidance (cwnd >= ssthresh)

23

Slow Start

• Initial value:– cwnd = 1 segment

• Note: cwnd is actually measured in bytes: 1 segment = MSS bytes

• Each time an ACK is received, the congestion window is increased by MSS bytes.

– cwnd = cwnd + MSS – If an ACK acknowledges two segments, cwnd is still increased by only 1

segment.

– Even if ACK acknowledges a segment that is smaller than MSS bytes long, cwnd is increased by MSS.

• Does Slow Start increment slowly? Not really. In fact, the increase of cwnd can be exponential

24

Slow Start Example

• The congestion window size grows very rapidly– For every ACK, we

increase cwnd by 1 irrespective of the number of segments ACK’ed

• TCP slows down the increase of cwnd when cwnd > ssthresh

cwnd =1xMSS

cwnd =2xMSS

cwnd =4xMSS

cwnd =7xMSS

25

Congestion Avoidance

• Congestion avoidance phase is started if cwnd has reached the slow-start threshold value

• If cwnd >= ssthresh then each time an ACK is received, increment cwnd as follows:

• cwnd = cwnd + MSS(MSS/ cwnd)

• So cwnd is increased by one segment (=MSS bytes) only if all segments have been acknowledged.

26

Slow Start / Congestion Avoidance

• Here we give a more accurate version than in our earlier discussion of Slow Start:

If cwnd <= ssthresh then Each time an Ack is received:cwnd = cwnd + MSS

else /* cwnd > ssthresh */

Each time an Ack is received :cwnd = cwnd + MSS. MSS / cwnd

endif

27

Example of Slow Start/Congestion Avoidance

Assume that ssthresh = 8 cwnd = 1

cwnd = 2

cwnd = 4

cwnd = 8

cwnd = 9

cwnd = 10

0

2

4

6

8

10

12

14

Roundtrip times

Cw

nd

(in

seg

men

ts)

ssthresh

28

Responses to Congestion

• Most often, a packet loss in a network is due to an overflow at a congested router (rather than due to a transmission error)

• So, TCP assumes there is congestion if it detects a packet loss

• A TCP sender can detect lost packets via:• Timeout of a retransmission timer• Receipt of a duplicate ACK

• When TCP assumes that a packet loss is caused by congestion it reduces the size of the sending window

29

TCP Tahoe

• Congestion is assumed if sender has timeout or receipt of duplicate ACK

• Each time when congestion occurs, – cwnd is reset to one:

cwnd = MSS– ssthresh is set to half the current size of the congestion

window:ssthressh = cwnd / 2

– and slow-start is entered

30

Slow Start / Congestion Avoidance

• A typical plot of cwnd for a TCP connection (MSS = 1500 bytes) with TCP Tahoe:

31

TCP Error Control

Background on Error Control

TCP Error Control

32

Background: ARQ Error Control

• Two types of errors:– Lost packets– Damaged packets

• Most Error Control techniques are based on:

1. Error Detection Scheme (Parity checks, CRC).

2. Retransmission Scheme.

• Error control schemes that involve error detection and retransmission of lost or corrupted packets are referred to as Automatic Repeat Request (ARQ) error control.

33

Background: ARQ Error Control

All retransmission schemes use all or a subset of the following procedures:

Positive acknowledgments (ACK) Negative acknowledgment (NACK) All retransmission schemes (using ACK, NACK or both) rely on the use of timers

The most common ARQ retransmission schemes are:Stop-and-Wait ARQ Go-Back-N ARQSelective Repeat ARQ

34

Background: ARQ Error Control

• The most common ARQ retransmission schemes:

– Stop-and-Wait ARQ

– Go-Back-N ARQ

– Selective Repeat ARQ

• The protocol for sending ACKs in all ARQ protocols are based on the sliding window flow control scheme

35

Background: Stop-and-Wait ARQ

• Stop-and-Wait ARQ is an addition to the Stop-and-Wait flow control protocol:

• Packets have 1-bit sequence numbers (SN = 0 or 1)• Receiver sends an ACK (1-SN) if packet SN is correctly

received• Sender waits for an ACK (1-SN) before transmitting the next

packet with sequence number 1-SN • If sender does not receive anything before a timeout value

expires, it retransmits packet SN

36

Background: Stop-and-Wait ARQ

Packet 1

• Lost Packet

A

B

AC

K 0

Packet 0

Timeout

AC

K 1

Packet 1

Packet 1 A

CK

0

37

Background: Go-Back-N ARQ

Operations:

– A station may send multiple packets as allowed by the window size

– Receiver sends a NAK i if packet i is in error. After that, the receiver discards all incoming packets until the packet in error was correctly retransmitted

– If sender receives a NAK i it will retransmit packet i and all packets i+1, i+2,... which have been sent, but not been acknowledged

38

Example of Go-Back-N ARQ

BA2

1

3

2

packetsreceived

packets waitingfor ACK/NAK

3

ACK2

packet 1 is received, send ACK 2

BA3

2

4

3

4

1

1

2

4

3

BA24

Time out for Packet 2retransmit frame 2,3,4

13

• In Go-back-N, if packets are correctly delivered, they are delivered in the correct sequence

• Therefore, the receiver does not need to keep track of `holes’ in the sequence of delivered packets

39

Background: Go-Back-N ARQ

Packet 0

• Lost Packet

A

B

Packet 2

Packet 4

Packet 1

Packet 3

AC

K 3

Packet 5

Packet 6

Packet 4

Packet 5

Packet 6

Packets 4,5,6are

retransmitted

AC

K 6

Packets 5 and 6are discarded

Timeoutfor Packet 4

40

Background: Selective-Repeat ARQ

• Similar to Go-Back-N ARQ. However, the sender only retransmits packets for which a time-out occured is received

• Advantage over Go-Back-N: – Fewer Retransmissions.

• Disadvantages: – More complexity at sender and receiver

– Each packet must be acknowledged individually (no cumulative acknowledgements)

– Receiver may receive packets out of sequence

41

Example of Selective-Repeat ARQ

BA2

1

3

2

Framesreceived

Packets waitingfor ACK/NAK

3

ACK2Packet is correct, send ACK 2Packet 2 does not arrive

BA3

2

4

3

4

ACK4

Following packets are Acked

1

1

2

4

3

BA2

4Timeout for packet 2:retransmit only packet2

1

3

5

5

ACK5

Receiver must keep track of `holes’ in the sequence of delivered packets

Sender must maintain one timer per outstanding packet

42

Timeout for Packet 4:only Packet 4

is retransmitted

Background: Selective-Repeat ARQ

Packet 0

• Lost Packet

A

B

Packet 2

Packet 1

Packet 3

AC

K 2

Packet 5

Packet 6

AC

K 6

Packet 4

Packet 7

Packet 0

AC

K 7

AC

K 1

AC

K 3

AC

K 4

Packet 4

Packets 5 and 6are buffered

AC

K 5

AC

K 0

AC

K 1

43

Error Control in TCP

• TCP implements a variation of the Go-back-N retransmission scheme

• TCP maintains a Retransmission Timer for each connection:– The timer is started during a transmission. A timeout

causes a retransmission

• TCP couples error control and congestion control (i.e., it assumes that errors are caused by congestion)

• TCP allows accelerated retransmissions (Fast Retransmit)

44

TCP Retransmission Timer

• Retransmission Timer:– The setting of the retransmission timer is crucial for

efficiency– Timeout value too small results in unnecessary

retransmissions– Timeout value too large long waiting time before

a retransmission can be issued

– A problem is that the delays in the network are not fixed – Therefore, the retransmission timers must be adaptive

45

Round-Trip Time Measurements

• The retransmission mechanism of TCP is adaptive • The retransmission timers are set based on round-trip time

(RTT) measurements that TCP performs

RTT #1

RTT #2

RTT #3

The RTT is based on time difference between segment transmission and ACKBut:

TCP does not ACK each segmentEach connection has only one timer

46

Round-Trip Time Measurements

• Retransmission timer is set to a Retransmission Timeout (RTO) value.

• RTO is calculated based on the RTT measurements. • The RTT measurements are smoothed by the following

estimators srtt and rttvar:

srttn+1 = RTT + (1- ) srttn

rttvarn+1 = ( | RTT - srttn+1 | ) + (1- ) rttvarn

RTOn+1 = srttn+1 + 4 rttvarn+1

• The gains are set to =1/4 and =1/8

• srtt0 = 0 sec, rttvar0 = 3 sec, Also: RTO1 = srtt1 + 2 rttvar1

47

Karn’s Algorithm

• If an ACK for a retransmitted segment is received, the sender cannot tell if the ACK belongs to the original or the retransmission.

Timeout !

RT

T ? R

TT

?

Karn’s Algorithm:Don’t update srtt on any segments that have been retransmitted.Each time when TCP retransmits, it sets:RTOn+1 = min( 2 RTOn, 64) (exponential backoff)

48

Measuring TCP Retransmission Timers

•Transfer file from Argon to neon

• Unplug Ethernet of Argon cable in the middle of file transfer

argon.tcpip-lab.edu("Argon")

neon.tcpip-lab.edu("Neon")

Transfer file

Web client Web server

49

Interpreting the Measurements

• The interval between retransmission attempts in seconds is:

1.03, 3, 6, 12, 24, 48, 64, 64, 64, 64, 64, 64, 64.

• Time between retrans-missions is doubled each time (Exponential Backoff Algorithm)

• Timer is not increased beyond 64 seconds

• TCP gives up after 13th attempt and 9 minutes.

0

100

200

300

400

500

600

Se

con

ds

0 2 4 6 8 10 12Transmission Attempts