Wide Area Networking Lecture 4: TCP protocol

1
Wide Area NetworkingLecture 4 TCP protocol
2
Transport layer - TCP

• Connection-oriented reliable byte-stream service
  over connectionless and unreliable IP
• More complex transmitter receiver
• Connection-oriented full-duplex connection
  between client server processes
• Connection setup, connection state, connection
  release
• Higher header overhead
• Error control, flow control, and congestion
  control
• Higher delay than UDP
• Applications using TCP
• HTTP, SMTP, FTP, TELNET, POP3, VoIP signalling
  protocols, ...

3
TCP operation principles

• TCP splits application data into the so-called
  segments, numbers them and sends
• Received segments have to be acknowledged by the
  receiver, otherwise segments are retransmitted
• Receiver reassembles the segments by numbers
• Error recovery
• Missed or corupted segments are retransmited
• Timeout mechanism to detect lost segments
• TCP deploys flow/congestion control mechanisms
  for better utilization of network bandwidth

4
TCP

• Connection-oriented
• Byte stream service
• Application sends byte blocks
• TCP forms them into segments transported in IP
  packets
• Receiving application reads received bytes

• Duplex connection
• Flow control Receiver controls rate at which
  sender transmits to prevent buffer overflow
• Congestion control Transmitter dynamically
  adjusts the rate according to congestion as
  indicated by RTT (round trip time) ACKs

IP packet
TCP segment
TCP data
TCP header20 bytes
IP header20 bytes
5
TCP

• TCP segment header

6
TCP

• Port Numbers
• A socket is a common name for IP address paired
  with a port number
• A connection is fully specified by a socket pair
• Well-known ports FTP 20 Telnet 23
• DNS 53 HTTP 80
• Sequence Number
• 32 bits long byte count with initial sequence
  number selected during connection setup
• Acknowledgement Number
• SN of next byte expected by receiver
• Acknowledges that all prior bytes in stream have
  been received correctly
• Valid if ACK flag is set
• Header length
• 4 bits length of header in multiples of 32-bit
  words
• Min. header length is 20 bytes, max. is 60 bytes

7
TCP

• Control (6 bits)
• URG urgent pointer flag
• Urgent message end SN urgent pointer
• ACK ACK packet flag
• PSH override TCP buffering
• RST reset connection
• Upon receipt of RST, connection is terminated and
  application layer notified
• SYN establish connection
• FIN close connection

8
TCP

• Window Size
• 16 bits to advertise window size
• Used for flow control
• Sender will accept bytes with SN from ACK to ACK
  window
• Maximum window size is 65535 bytes
• TCP Checksum
• Internet checksum method
• TCP pseudo header TCP segment
• Options
• Variable length
• NOP (No Operation) option is used to pad TCP
  header to multiple of 32 bits
• Timestamp option is used for round trip
  measurements
• Maximum Segment Size (MSS) option specifies
  largest segment a receiver wants to receive
• Window Scale option increases TCP window from 16
  to 32 bits

9
TCP Connection management

• Connection establishment Three-way Handshake

10
TCP connection management

• Connection teardown

11
TCP States
12
Flow control

• Sliding window mechanism
• a window determines a number of bytes that can be
  injected to the network without acknowledgement

receiver window
current window
1
2
3
4
5
6
7
8
9
10
11
...
sent but not acknowledged
have to wait for available window
sent and acknowledged
can be sent immediately
13
TCP Congestion Control

• Advertised window size
• used to ensure that receivers buffer will not
  overflow
• However, buffers at intermediate routers between
  source and destination may overflow
• Congestion occurs when total arrival rate from
  all packet flows exceeds R over a sustained
  period of time
• Buffers at multiplexer will fill and packets will
  be lost

R
Throughput (bps)
Arrival Rate
Delay (sec)
Arrival Rate
R
14
Window Congestion Control

• Window mechanism has a desired operating point
• Sources must control their sending rates so that
  aggregate arrival rate is just before the
  congestion point
• TCP sender maintains a congestion window cwnd to
  control congestion at intermediate routers
• Effective window is minimum of congestion window
  and advertised window
• Problem source does not know what its fair
  share of available bandwidth should be
• Solution adapt dynamically to available BW
• Sources probe the network by increasing cwnd
• When congestion detected, sources reduce rate
• Ideally, sources sending rate stabilizes near
  ideal point

15
Congestion Window

• TCP congestion algorithm changes congestion
  window dynamically according to the network state
• At light traffic each segment is ACKed quickly
• Increase cwnd aggressively
• Near congestion point segment ACKs arrive, but
  more slowly
• Slow down increase in cwnd
• At congestion segments encounter large delays
  (so retransmission timeouts occur) segments are
  dropped in router buffers (resulting in duplicate
  ACKs)
• Reduce transmission rate, then probe again

16
TCP Congestion Control Slow Start

• Slow start increase congestion window size by
  one segment upon receiving an ACK from receiver
• initialized at ? 2 segments
• used at (re)start of data transfer
• congestion window increases exponentially

Seg
ACK
17
TCP Congestion Control Congestion Avoidance

• Algorithm progressively sets a congestion
  threshold
• When cwnd gt threshold, slow down rate at which
  cwnd is increased
• Increase congestion window size by one segment
  per round-trip-time (RTT)
• Each time an ACK arrives, cwnd is increased by
  1/cwnd
• In one RTT, cwnd segments are sent, so total
  increase in cwnd is cwnd x 1/cwnd 1
• cwnd grows linearly with time

18
TCP Congestion Control Congestion

• Congestion is detected upon timeout or receipt of
  duplicate ACKs
• Assume current cwnd corresponds to available
  bandwidth
• Adjust congestion threshold ½ x current cwnd
• Reset cwnd to 1
• Go back to slow-start
• Over several cycles expect to converge to
  congestion threshold equal to about ½ the
  available bandwidth

19
Fast Retransmit Fast Recovery

• Congestion causes many segments to be dropped
• If only a single segment is dropped, then
  subsequent segments trigger duplicate ACKs before
  timeout
• Can avoid large decrease in cwnd as follows
• When three duplicate ACKs arrive, retransmit lost
  segment immediately
• Reset congestion threshold to ½ cwnd
• Reset cwnd to congestion threshold 3 to account
  for the three segments that triggered duplicate
  ACKs
• Remain in congestion avoidance phase
• However if timeout expires, reset cwnd to 1
• In absence of timeouts, cwnd will oscillate
  around optimal value

SN1
ACK2
SN2
SN3
SN4
ACK2
SN5
ACK2
ACK2
20
TCP Congestion Control Fast Retransmit Fast
Recovery
Congestion avoidance
20
Time-out
15
Threshold
Congestion window
10
Slow start
5
0
Round-trip times
21
TCP observations

• TCP connection throughput sawtooth curve
• TCP is an adaptive or feedback-based protocol
  it will try to exploit available bandwidth as
  much as possible (until limitations arise)

22
Delayed ACK

• Sending empty ACK packets (only ACK, no data)
  consumes network bandwidth
• Ineffective network utilisation
• Short packet problem
• Delayed ACK
• Receiver waits up to 200 ms to send ACK if it has
  no data to send
• Every second segment is acknowledged without
  waiting for delay ACK timeout
• When ACK timeout expires the acknowledgement is
  sent hopefully with data that was buffered
  meantime

23
Nagles Algorithm

• Very short datagrams (tinygrams) consume network
  resource ineffectively
• Nagles algorithms mitigates this problem
• Only one short segment can be outstanding for
  acknowledgement
• While waiting for acknowledgement the arriving
  (from upper layer) data is buffered
• When acknowledgement arrive or there is enough
  data to fill the packet the next packet can be
  send
• Nagles algorithm initially buffers some data and
  then starts sending packets
• Nagles algorithm automatically adopts to the
  network state (RTT)
• Short RTT- no influence
• Large RTT more data will be buffed, larger
  initial delay
• No short packets - no influence

24
Nagles algorithm example

• (1) first segment is send immediately
• (2) the arriving data is buffered
• (3) when the acknowledgement arrives all buffered
  data is send (up to window size)
• (4) in the steady state the efficiency of the TCP
  connection with Nagles algorithm is the same as
  original TCP protocol

25
Timeout and retransmission

• The reliable data transfer is ensured by the
  acknowledgement and retransmission mechanism
• After sending the segment the timer is started
• When the ACK is not received before timeout
  expires the segments (starting from the last
  acknowledged segment) are automatically
  retransmitted
• The timeout (RTO) value has significant impact on
  the TCP performance
• Too short RTO unnecessary retransmission
  network overload, slow start triggering, low
  throughput
• Too large RTO long inactivity periods low
  throughput
• How to set the value of RTO
• Adaptive measurement procedure

26
RTO measurement (1)

• The RTO is measured based on the RTT (Round Trip
  Time) estimation
• The average RTT is estimated by so called
  weighted moving average
• R aR(1-a)M
• R - estimated average value of RTT M - the RTT
  value from last measurement
• a estimator weight
• The RTO (timeout) value is given by
• RTO Rb
• b - is the assumed measure of RTT variance
  (equal 2).
• The above procedure does not work well when the
  RTT fluctuations are high

27
RTO measurement (1)

• To improve the RTO measurement the variance of
  the RTT must be taken into account
• The average RTT is estimated by the following
  procedure
• Err M - A
• A A gErr
• A - estimated average value of RTT M - the RTT
  value from last measurement
• g estimator weight 0.125
• The variance of RTT is estimated by the following
  procedure
• D D h(Err - D)
• h - algorithm parameter 0,25
• The RTO (timeout) value is given by
• RTO A 4D

28
Karns Algorithm (1)

• RTO measurement uncertainty
• During retransmission we cannot distinguish the
  following situations
• (A) measured RTT is related to originally
  transmitted segment
• (B) measured RTT is related to retransmitted
  segment
• Using rule (A) will increase RTO and decrease the
  TCP throughput in case of transmission errors
• i.e. on wireless links
• Using rule (B) will increase the number of
  unnecessary retransmission again lowering TCP
  throughput

Segment
RTO
retransmission
(A)
Ack
(B)
What is true (A) or (B)?
29
Karns Algorithm (2)

• Use only these RTT values that are not related to
  retransmission
• Shall retransmission occur, double the RTO value
  (up to maximum value of 64 sec.) - exponential
  backoff
• The protection against keeping too short values
  of RTO for long time

30
Useful equation

• ThroughputltTCP Window Size/RTT
• TCP Window SizegtBandwidthRTT
• Buffer Size gt 2BandwidthRTT
• After 3BandwidthRTT data has been send the TCP
  will reach the optimal performance
• Throughput 0,7MTU/RTT/SqrtPloss