Internet Engineering 2018 Transport layer: TCPFrom byte-stream to packets: TCP header IP Header TCP...

Internet Engineering 2018Transport layer: TCP

Youki KadobayashiNAIST

Copyright(C)2018 Youki Kadobayashi. All rights reserved.

ApplicationPresentation

SessionTransportNetwork

Data LinkPhysical

ApplicationPresentation

SessionTransportNetwork

Data LinkPhysical

2

Transport layer: a birds-eye view

H HRRR R

IPIP IP IPIP

Transport

Routers don’t maintain per-host state

Hosts maintain state for each transport-layer endpoint


Functions provided by the transport layer

n Communication between processes¨ Identify process¨ Identify inter-process channel

n Interface for upper layer¨ Connection-oriented (virtual circuit)¨ Connectionless (datagram)

n Competition and arbitration of network resource¨ Flow control¨ Congestion control

Next lecture

P Q

(P, Q)

3


Transport protocols in the Internet protocol suite

n TCP (RFC793)¨ Transmission Control Protocol

¨ Connection-oriented¨ Multiple functions

n for reliability

n SCTP (RFC4960)

n UDP (RFC768)¨ User Datagram Protocol

¨ Connectionless¨ IP & Process identification

n Packets can be lost

n DCCP (RFC4340)Advanced topic

4


Identifying process and connection in TCP

n Unique identification of process¨(IP, port)

n Unique identification of TCP connection¨ (source IP, source port, destination IP, destination port)

1040

80

22

2137

(203.178.136.36, 22)

163.221.52.100 203.178.136.36

(163.221.52.100, 1040)

connection

connection

process

5


Some well-known protocols/ports

Port Protocol Use20,21 FTP File transfer

22 SSH Remote login, replacement for telnet23 TELNET Telnet25 SMTP Email 80 HTTP World Wide Web

110 POP-3 Remote email access143 IMAP Remote email access443 HTTPS Secure Web (HTTP over SSL/TLS)543 RTSP Media player control631 IPP Printer sharing

6https://www.iana.org


TCP service model (1)

n Connection-orientedn Virtual Circuit

¨ Looks like a circuit, but virtual: no physical wires between specific endpoints

n Adapts to speed¨ Adapts to speed of intermediate networks as well as the

speed of endpoints

7


Establishing a TCP connection

n 3-way handshaken SYN, SYN-ACK, ACKn Ensure full-duplex communication

16bit source port 16bit destination port

32bit sequence number

32bit acknowledgment number

4bit hlen 16bit window size

16bit TCP checksum 16bit urgent pointer

reserved flags

URG ACK PSH RST SYN FIN

SYN

SYN-ACK

ACK

8


Establishing TCP connection: a concrete example with Wiresharkn # tshark -i en0 -n -f 'port 80'n tcpdump: listening on de0n Capturing on en0n 0.164720 10.0.1.148 -> 163.221.8.221 TCP 63428 > 80 [SYN] Seq=0

Win=65535 Len=0 MSS=1460 WS=16 TSval=1135377929 TSecr=0 SACK_PERM=1

n 0.195154 163.221.8.221 -> 10.0.1.148 TCP 80 > 63428 [SYN, ACK] Seq=0 Ack=1 Win=50137 Len=0 TSval=190559467 TSecr=1135377929 MSS=1460 WS=8 SACK_PERM=1

n 0.195322 10.0.1.148 -> 163.221.8.221 TCP 63428 > 80 [ACK] Seq=1 Ack=1 Win=131760 Len=0 TSval=1135377958 TSecr=190559467

Reply with Sequence number + 1 as an ACK implies acknowledgment

9


Meanings of Wireshark output

n time source IP -> dest IP TCP source port > dest port [ flags ] Seq=n Ack=n …

n 0.195154 163.221.8.221 -> 10.0.1.148 TCP 80 > 63428 [SYN, ACK] Seq=0 Ack=1 Win=50137 Len=0 TSval=190559467 TSecr=1135377929 MSS=1460 WS=8 SACK_PERM=1

10


Virtual Circuit(2):TCP connection release

FIN

Ack of FIN

FIN

Ack of FIN

close

close

5.749366 163.221.8.221 -> 10.0.1.148 TCP 80 > 63428 [FIN, ACK] Seq=659 Ack=2449 Win=401096 Len=0 TSval=190560021 TSecr=11353784825.749464 10.0.1.148 -> 163.221.8.221 TCP 63428 > 80 [ACK] Seq=2449 Ack=660 Win=131104 Len=0 TSval=1135383482 TSecr=1905600215.749650 10.0.1.148 -> 163.221.8.221 TCP 63428 > 80 [FIN, ACK] Seq=2449 Ack=660 Win=131104 Len=0 TSval=1135383482 TSecr=1905600215.765279 163.221.8.221 -> 10.0.1.148 TCP 80 > 63428 [ACK] Seq=660 Ack=2450Win=401096 Len=0 TSval=190560024 TSecr=1135383482

11


TCP connection reset

n RST¨ Abortive release¨ Nonexistent port (e.g., dead process)

telnet to sh.naist.jp, port 80 will result in connection reset:n 0.000000 10.0.1.148 -> 163.221.10.10 TCP 63436 > 80 [SYN] Seq=0 Win=65535

Len=0 MSS=1460 WS=16 TSval=1137289324 TSecr=0 SACK_PERM=1n 0.018841 163.221.10.10 -> 10.0.1.148 TCP 80 > 63436 [RST, ACK] Seq=1

Ack=1 Win=0 Len=0

If you kill “Dropbox”:n 55.992214 10.0.1.148 -> 108.160.163.51 TCP 62991 > 80 [RST] Seq=310 Win=0

Len=012


Virtual Circuit Summary:TCP state transition diagram

State transition:trigger / response

13


$ netstatActive Internet connectionsProto Recv-Q Send-Q Local Address Foreign Address (state)tcp4 0 0 45.1.20.101.57456 74.125.235.138.http SYN_SENTtcp4 0 0 45.1.20.101.57455 ey-in-f101.1e100.http ESTABLISHEDtcp4 0 0 45.1.20.101.57454 74.125.235.148.http ESTABLISHED

Troubleshooting TCP with state machine

n netstat

SYN sent, awaiting SYN+ACK response

Many other open-source tools are available:lsof, trpt, tcptraceroute, tcptrace, tcpflow, etc.Try some of these tools in the provided VM image.

14


Hands-on: observe connection setup/release

In the provided virtual machine,1. Observe connection setup/release;

ë.g., by using browser within VM2. Observe connection reset by terminating program;

ë.g., by skill firefox3. Observe connection reset by connecting to

nonexistent port on the working machine.ë.g., by telnet sh.naist.jp 80

15


Buffered transfer: adapts to varying speed of endpoints and intermediate networks

TCP connection

sendbuffer

recvbuffer

sendbuffer

recvbuffer

Process

OS kernel

block/unblockRead()Write()

Process

Read()Write()

OS kernel16


TCP service model (2)

n Byte-stream service¨ Upper layer can’t see boundaries between packets¨ no boundary: structuring (framing) is needed at upper layer

n Full duplex¨ independent two streams in single connection

n Reliable¨ masks packet reordering, duplication, discard and bit error

O L L E H O L L E HTCP being viewed asbyte-stream serviceOK OK

17


How does TCP implement reliable stream service?

n ACK: Acknowledgment¨ Active acknowledgment

n Explicitly acknowledge the receipt of packet¨ Duplicate ACK

n Implicitly communicate the packet loss information

n Timeout and Retransmission¨ Whenever sender doesn’t receive ACK after fixed time,

a “Timeout” event is triggeredp Retransmission: assuming that transmission has failed

n Exponential back-off: 3, 6, 12, …

18


ACK: Acknowledgment

Packets in transit

Sent but unacknowledgedSent and acknowledged

User data arrives

Sender

Receiver

Nara Institute of Science and Technology

Nara Insti

10

16

19


Piggybacking: Exploiting full-duplex channel

Packets in transit


User data arrives

SenderReceiver

ReceiverSender


Nara Insti

Graduate School of Information Science

Graduate S

User data arrives

Note: rough sketch

Sent but unacknowledgedSent and acknowledged 20


Duplicate ACK: implicit communication of packet loss

Packets in transit


User data arrives

Sender

Receiver


Nara Institute o

Packet loss

10

16

16

21


From byte-stream to packets: TCP header

IPHeader

TCPHeader TCP data

TCP segment

16bit source port 16bit destination port

32bit sequence number

32bit acknowledgment number

4bit hlen 16bit window size

16bit TCP checksum 16bit urgent pointer

(options)

(TCP data)

reserved flags

20octets

Chop appropriate length from byte-stream buffer & add TCP header

22


Nagle algorithm

n Q. If you added 20byte+20byte size header to 1byte data, isn’t that overhead big?

n Nagle algorithm (RFC896)¨ There is only one small segment which is unacknowledged

in the network.¨ In case of short RTT:à acceptable overhead, as LAN bandwidth is abundantà send packets with small buffering

¨ In case of long RTTà reduce overhead, due to WAN bandwidth constraints

23


Hands-on: observe packet loss & retransmission

Within provided virtual machine,1.Observe duplicate ACK;

¨Emulate lossy network with:¨ # tc qdisc add dev eth1 root netem loss 10%

1.Observe the effect of Nagle algorithm;¨Emulate satellite network with:¨ # tc qdisc add dev eth1 root netem delay 500ms

24


Questions?

25


Summary thus far

n Transport layer

n The transport protocol on the internet – TCPn TCP: service model, and features

n Efficiency: ACK, piggybacking, Nagle algorithmn TCP connection establishment and releasen Diagnosis: tools + state machine knowledge

26


Transport protocol challenges

n With a number of unknown parameters:¨ Number of active communications¨ Bottleneck bandwidth¨ Error rate

n how can we accommodate as much communications as possible, without collapsing network itself?

1 1

n k

… …

Key ideas: probing, estimation, self-policing, macro-level stability27


Flow control and congestion control:Competition and Arbitration of Network Resource

n Flow control¨Absorb difference of transmission rate¨Recovery from sequence error¨Recovery from duplication, packets drop and bit

errorn Congestion control

¨Sharing the bandwidth while suppressing the congestion

¨Fair sharing of bandwidth

28


Characteristics of TCP Flow Control

n End to end¨ No global assignment of resource¨ Estimate available bandwidth at individual hosts¨ Routers do not explicitly allocate resource

n Implicit signaling through packet drops

n Scalable¨ Host-based autonomous method has better scalability,

because it doesn’t need state management inside network.n Autonomous à Less state management à Scalable

29


TCP: key characteristics

n Very simple algorithm¨ Macroscopic self-stabilization

n TCP doesn't assume the presence of greedy nodes.¨ Eliminate global control system¨ Reject the idea of intermediate policing system

n Modest performance across almost all data-links¨ .. as opposed to optimal performance in specific condition

n Stepwise improvements over 30 years

30


Flow Control on TCP

n Control available bandwidth¨ Sliding window

n Using sequence number, not packet number¨ Window size

n Control transmission interval of packets¨ ACK clocking

n Miscellaneous¨ Error detection with TCP checksum¨ Packet drop detection with duplicate ACK, timeout

31

2009-08/congavoid.pdf


Sliding window

Packets in flight


User data arrives

Sender

Receiver


Nara Insti

10

16

Window size Sequence number

32


Congestion Control on TCP

n Fair-share model: fair distribution of bandwidthEnd to end

n Increase and decrease of window size¨ additive increase¨ multiplicative decrease

n AIMD is known to be self-stabilizing (R. Jain, et al., “Analysis of the increase and decrease algorithms for congestion avoidance in computer networks”, 1989)

n Switch increasing strategy of window sizen Detect congestion through packet drops

33


Increasing Window size

n TCP switches increasing strategy of congestion window (cwnd) by slow start threshold (ssthresh)

(Outline of the algorithm)n On receiving an ACK:

if (cwnd < ssthresh) {/* slow start: exponential increase */cwnd += 1;

} else {/* congestion avoidance: additive increase

*/cwnd += 1 / cwnd;

}

34


Increasing Window size

n Slow start¨ exponential increase

n Congestion avoidance¨ additive increase

n Effectiveness:see V. Jacobson, “Congestion Avoidance and Control”, SIGCOMM’88.

Source: TCP/IP Illustrated, Vol.1

35


Decreasing Window size

(Outline of algorithm )n On detecting packet drop:

ssthresh = cwnd / 2;if (timeout) {

cwnd = 1;}


36


Hands-on: observe slow start & congestion avoidance

n Start Wiresharkn Access speed test website:

¨ http://www.speedtest6.com

n Filter the TCP stream with:¨ ip.src==27.120.100.45

n Plot the Time-Sequence Graph¨ Select one of the filtered packets¨ Go to Statistics -> TCP StreamGraph -> Time Sequence Graph

(Stevens)¨ Go to Statistics -> TCP StreamGraph -> Time Sequence Graph

(tcptrace)

37


Questions?

38


Packet drop

n Detection with 2 methods:¨Duplicate ACK¨Retransmission Time Out (RTO)

n How do I judge the time out?→ RTO ~ RTT Estimator

39


RTO Calculation

n Err = M – AA <- A + gErrD <- D + h(|Err| - D)RTO = A + 4D

¨A: smoothed RTT¨D: smoothed mean deviation¨g: gain for the average (1/8)¨h: gain for the deviation (1/4)


40


RTO calculation

Source: A Quick Tour Around TCP,http://web.eecs.utk.edu/~dunigan/tcptour/

41


User Datagram Protocol (UDP) RFC 768

n “Unreliable Data Protocol” (J)n Connectionless transport protocol

¨ Avoid overhead and delays of ordered, reliable delivery¨ Send messages to and receive them from a socket

n Lightweight delivery service¨ IP and port numbers to support (de)multiplexing¨ Optional error checking on the packet contents

42


Incentives on using UDP

n Finer control over what data is sent and when¨ UDP packages the data and send the packet as soon as an

application process writes into the socket

n No delay for connection establishment¨ There are no preliminaries in UDP

n No connection state¨ No allocation of buffers, parameters, sequence numbers,

etc.

n Small packet header overhead: only eight -bytes long

43


UDP applications

n Multimedia streaming¨ Real-time Transport Protocol (RTP): internet radio,

telephony, music-on-demand, video-on-demand¨ Retransmission of lost/corrupted packets is not worthwhile

n Simple query protocols like Domain Name System¨ Overhead of connection establishment is overkill¨ Easier to have application retransmit if needed

44


Summary

n Transport layern The transport protocol on the internet – TCPn TCP: service model, and features

n Efficiency: ACK, piggybacking, Nagle algorithmn TCP connection establishment and releasen Flow control and congestion control in TCP

45


Assignment 5: Observe TCP mechanisms in action

n Use iperf3 to establish TCP connections with four servers provided by our TA and capture the TCP traffic between the servers and the client.¨1) Explain details of its TCP connection.¨2) (optional) Cause delay and packet loss in the

client side by using tc command, then conduct the same investigation as above.

n Submission deadline: July 24, 17:00 JST

46


Hands-on: observe the effect of loss/delay

n Start Wiresharkn Start iperf

¨ Install iperf3 on the VM: $ sudo apt install iperf3¨ run iperf3: $ iperf3 -c 163.221.52.133 -p 5201 -t 2

n Configure tc¨ Loss: $ sudo tc qdisc add dev eth1 root netem loss 10%

n To confirm: $ sudo tc qdisc shown To remove: $ sudo tc qdisc del dev eth1 root netem loss 10%

¨ Delay: $sudo tc qdisc add dev eth1 root netem delay 500ms

n Observe the effect of loss/delay¨ Select one of the filtered packets¨ Go to Statistics -> TCP StreamGraph -> Time Sequence Graph (Stevens)¨ Go to Statistics -> TCP StreamGraph -> Time Sequence Graph (tcptrace)

47

Internet Engineering 2018 Transport layer: TCPFrom byte-stream to packets: TCP header IP Header TCP...

Documents

Transcript of Internet Engineering 2018 Transport layer: TCPFrom byte-stream to packets: TCP header IP Header TCP...