2. Transport Layer 3-2
Chapter 3: Transport Layer
our goals:
īļ understand
principles behind
transport layer
services:
ī§ multiplexing,
demultiplexing
ī§ reliable data
transfer
ī§ flow control
ī§ congestion control
īļ learn about Internet
transport layer
protocols:
ī§ UDP: connectionless
transport
ī§ TCP: connection-
oriented reliable
transport
ī§ TCP congestion control
3. Transport Layer 3-3
Chapter 3 outline
3.1 transport-layer
services
3.2 multiplexing and
demultiplexing
3.3 connectionless
transport: UDP
3.4 principles of
reliable data
transfer
3.5 connection-oriented
transport: TCP
ī§ segment structure
ī§ reliable data transfer
ī§ flow control
ī§ connection
management
3.6 principles of
congestion control
3.7 TCP congestion
control
4. Transport Layer 3-4
Transport services and protocols
īļ provide logical
communication between
app processes running on
different hosts
īļ transport protocols run in
end systems
ī§ send side: breaks app
messages into
segments, passes to
network layer
ī§ rcv side: reassembles
segments into
messages, passes to
app layer
īļ more than one transport
protocol available to apps
application
transport
network
data link
physical
application
transport
network
data link
physical
5. Transport Layer 3-5
Transport vs. network layer
īļ network layer:
logical
communication
between hosts
īļ transport layer:
logical
communication
between
processes
ī§ relies on,
enhances,
network layer
services
12 kids in Annâs house
sending letters to 12 kids
in Billâs house:
īļ hosts = houses
īļ processes = kids
īļ app messages = letters
in envelopes
īļ transport protocol = Ann
and Bill who demux to in-
house siblings
īļ network-layer protocol =
postal service
household analogy:
6. Transport Layer 3-6
Internet transport-layer protocols
īļ reliable, in-order
delivery (TCP)
ī§ congestion control
ī§ flow control
ī§ connection setup
īļ unreliable, unordered
delivery: UDP
ī§ no-frills extension of
âbest-effortâ IP
īļ services not
available:
ī§ delay guarantees
ī§ bandwidth guarantees
application
transport
network
data link
physical
application
transport
network
data link
physical
network
data link
physical
network
data link
physical
network
data link
physical
network
data link
physical
network
data link
physical
network
data link
physical
network
data link
physical
7. Transport Layer 3-7
Chapter 3 outline
3.1 transport-layer
services
3.2 multiplexing and
demultiplexing
3.3 connectionless
transport: UDP
3.4 principles of
reliable data
transfer
3.5 connection-oriented
transport: TCP
ī§ segment structure
ī§ reliable data transfer
ī§ flow control
ī§ connection
management
3.6 principles of
congestion control
3.7 TCP congestion
control
8. Transport Layer 3-8
Multiplexing/demultiplexing
process
socket
use header info to deliver
received segments to correct
socket
demultiplexing at receiver:
handle data from multiple
sockets, add transport header
(later used for demultiplexing)
multiplexing at sender:
transport
application
physical
link
network
P2P1
transport
application
physical
link
network
P4
transport
application
physical
link
network
P3
9. Transport Layer 3-9
How demultiplexing works
īļ host receives IP
datagrams
ī§ each datagram has
source IP address,
destination IP address
ī§ each datagram carries
one transport-layer
segment
ī§ each segment has
source, destination port
number
īļ host uses IP addresses
& port numbers to direct
segment to appropriate
socket
source port # dest port #
32 bits
application
data
(payload)
other header fields
TCP/UDP segment format
10. Transport Layer 3-10
Connectionless demultiplexing
īļ when host receives
UDP segment:
ī§ checks destination port
# in segment
ī§ directs UDP segment to
socket with that port #
īļ recall: when creating
datagram to send into
UDP socket, must
specify
ī§ destination IP address
ī§ destination port #
IP datagrams with
same dest. port #, but
different source IP
addresses and/or
source port numbers
will be directed to same
socket at dest
11. Transport Layer 3-12
Connection-oriented demux
īļ TCP socket identified
by 4-tuple:
ī§ source IP address
ī§ source port number
ī§ dest IP address
ī§ dest port number
īļ demux: receiver uses
all four values to
direct segment to
appropriate socket
īļ server host may
support many
simultaneous TCP
sockets:
ī§ each socket identified
by its own 4-tuple
īļ web servers have
different sockets for
each connecting client
ī§ non-persistent HTTP
will have different
socket for each request
12. Transport Layer 3-15
Chapter 3 outline
3.1 transport-layer
services
3.2 multiplexing and
demultiplexing
3.3 connectionless
transport: UDP
3.4 principles of
reliable data
transfer
3.5 connection-oriented
transport: TCP
ī§ segment structure
ī§ reliable data transfer
ī§ flow control
ī§ connection
management
3.6 principles of
congestion control
3.7 TCP congestion
control
13. Transport Layer 3-16
UDP: User Datagram Protocol [RFC
768]
īļ âno frills, Internet
transport protocol
īļ âbest effortâ service,
UDP segments may be:
ī§ lost
ī§ delivered out-of-order
to app
īļ connectionless:
ī§ no handshaking
between UDP
sender, receiver
ī§ each UDP segment
handled
independently of
others
īļ UDP use:
ī§ streaming multimedia
apps (loss tolerant,
rate sensitive)
ī§ DNS
ī§ SNMP
īļ reliable transfer over
UDP:
ī§ add reliability at
application layer
ī§ application-specific
error recovery!
14. Transport Layer 3-17
UDP: segment header
source port # dest port #
32 bits
application
data
(payload)
UDP segment format
length checksum
length, in bytes of
UDP segment,
including header
īļ no connection
establishment (which
can add delay)
īļ simple: no connection
state at sender, receiver
īļ small header size
īļ no congestion control:
UDP can blast away as
fast as desired
why is there a UDP?
15. Transport Layer 3-18
UDP checksum
sender:
īļ treat segment
contents, including
header fields, as
sequence of 16-bit
integers
īļ checksum: addition
(oneâs complement
sum) of segment
contents
īļ sender puts checksum
value into UDP
checksum field
receiver:
īļ compute checksum of
received segment
īļ check if computed
checksum equals
checksum field value:
ī§ NO - error detected
ī§ YES - no error
detected. But maybe
errors nonetheless?
More later âĻ.
Goal: detect âerrorsâ (e.g., flipped bits) in
transmitted segment
16. Transport Layer 3-19
Internet checksum: example
example: add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0
1 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 0
1 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sum
checksum
Note: when adding numbers, a carryout from the most
significant bit needs to be added to the result
17. Transport Layer 3-20
Chapter 3 outline
3.1 transport-layer
services
3.2 multiplexing and
demultiplexing
3.3 connectionless
transport: UDP
3.4 principles of
reliable data
transfer
3.5 connection-oriented
transport: TCP
ī§ segment structure
ī§ reliable data transfer
ī§ flow control
ī§ connection
management
3.6 principles of
congestion control
3.7 TCP congestion
control
18. Transport Layer 3-21
Principles of reliable data
transfer
īļ important in application, transport, link layers
ī§ top-10 list of important networking topics!
īļ characteristics of unreliable channel will determine
complexity of reliable data transfer protocol (rdt)
19. Transport Layer 3-22
īļ characteristics of unreliable channel will determine
complexity of reliable data transfer protocol (rdt)
Principles of reliable data
transfer
īļ important in application, transport, link layers
ī§ top-10 list of important networking topics!
20. Transport Layer 3-23
īļ characteristics of unreliable channel will determine
complexity of reliable data transfer protocol (rdt)
īļ important in application, transport, link layers
ī§ top-10 list of important networking topics!
Principles of reliable data
transfer
21. Transport Layer 3-25
weâll:
īļ incrementally develop sender, receiver sides
of reliable data transfer protocol (rdt)
īļ consider only unidirectional data transfer
ī§ but control info will flow on both directions!
īļ use finite state machines (FSM) to specify
sender, receiver
state
1
state
2
event causing state transition
actions taken on state transition
state: when in this
âstateâ next state
uniquely determined
by next event
event
actions
Reliable data transfer: getting started
22. Transport Layer 3-26
rdt1.0: reliable transfer over a reliable
channel
īļ underlying channel perfectly reliable
ī§ no bit errors
ī§ no loss of packets
īļ separate FSMs for sender, receiver:
ī§ sender sends data into underlying channel
ī§ receiver reads data from underlying channel
Wait for
call from
above packet = make_pkt(data)
udt_send(packet)
rdt_send(data)
extract (packet,data)
deliver_data(data)
Wait for
call from
below
rdt_rcv(packet)
sender receiver
23. Transport Layer 3-28
īļ underlying channel may flip bits in packet
ī§ checksum to detect bit errors
īļ the question: how to recover from errors:
ī§ acknowledgements (ACKs): receiver explicitly tells
sender that pkt received OK
ī§ negative acknowledgements (NAKs): receiver
explicitly tells sender that pkt had errors
ī§ sender retransmits pkt on receipt of NAK
īļ new mechanisms in rdt2.0 (beyond rdt1.0):
ī§ error detection
ī§ feedback: control msgs (ACK,NAK) from receiver to
sender
rdt2.0: channel with bit errors
24. Transport Layer 3-32
rdt2.0 has a fatal flaw!
what happens if
ACK/NAK
corrupted?
īļ sender doesnât know
what happened at
receiver!
īļ canât just retransmit:
possible duplicate
handling duplicates:
īļ sender retransmits
current pkt if ACK/NAK
corrupted
īļ sender adds sequence
number to each pkt
īļ receiver discards
(doesnât deliver up)
duplicate pktstop and wait
sender sends one
packet,
then waits for receiver
response
25. Transport Layer 3-35
rdt2.1: discussion
sender:
īļ seq # added to pkt
īļ two seq. #âs (0,1)
will suffice. Why?
īļ must check if
received ACK/NAK
corrupted
īļ twice as many
states
ī§ state must
ârememberâ whether
âexpectedâ pkt
should have seq # of
0 or 1
receiver:
īļ must check if
received packet is
duplicate
ī§ state indicates
whether 0 or 1 is
expected pkt seq #
īļ note: receiver can
not know if its last
ACK/NAK received
OK at sender
26. Transport Layer 3-38
rdt3.0: channels with errors and loss
new assumption:
underlying channel
can also lose
packets (data,
ACKs)
ī§ checksum, seq. #,
ACKs,
retransmissions will
be of help âĻ but not
enough
approach: sender waits
âreasonableâ amount
of time for ACK
īļ retransmits if no ACK
received in this time
īļ if pkt (or ACK) just
delayed (not lost):
ī§ retransmission will be
duplicate, but seq. #âs
already handles this
ī§ receiver must specify
seq # of pkt being
ACKed
īļ requires countdown timer
27. Transport Layer 3-42
rdt3.0: stop-and-wait operation
first packet bit transmitted, t = 0
sender receiver
RTT
last packet bit transmitted, t = L / R
first packet bit arrives
last packet bit arrives, send ACK
ACK arrives, send next
packet, t = RTT + L / R
U
sender =
.008
30.008
= 0.00027
L / R
RTT + L / R
=
28. Transport Layer 3-43
Performance of rdt3.0
īļ rdt3.0 is correct, but performance stinks
īļ e.g.: 1 Gbps link, 15 ms prop. delay, 8000 bit
packet:
ī§ U sender: utilization â fraction of time sender busy
sending
U
sender =
.008
30.008
= 0.00027
L / R
RTT + L / R
=
ī§ if RTT=30 msec, 1KB pkt every 30 msec: 33kB/sec
thruput over 1 Gbps link
īļ network protocol limits use of physical resources!
Dtrans =
L
R
8000 bits
109 bits/sec
= = 8 microsecs
29. Transport Layer 3-56
Chapter 3 outline
3.1 transport-layer
services
3.2 multiplexing and
demultiplexing
3.3 connectionless
transport: UDP
3.4 principles of
reliable data
transfer
3.5 connection-oriented
transport: TCP
ī§ segment structure
ī§ reliable data transfer
ī§ flow control
ī§ connection
management
3.6 principles of
congestion control
3.7 TCP congestion
control
30. Transport Layer 3-57
TCP: Overview RFCs: 793,1122,1323, 2018,
2581
īļ full duplex data:
ī§ bi-directional data
flow in same
connection
ī§ MSS: maximum
segment size
īļ connection-oriented:
ī§ handshaking
(exchange of control
msgs) inits sender,
receiver state before
data exchange
īļ flow controlled:
ī§ sender will not
overwhelm receiver
īļ point-to-point:
ī§ one sender, one
receiver
īļ reliable, in-order byte
steam:
ī§ no âmessage
boundariesâ
īļ pipelined:
ī§ TCP congestion and
flow control set
window size
31. Transport Layer 3-58
TCP segment structure
source port # dest port #
32 bits
application
data
(variable length)
sequence number
acknowledgement number
receive window
Urg data pointerchecksum
FSRPAU
head
len
not
used
options (variable length)
URG: urgent data
(generally not used)
ACK: ACK #
valid
PSH: push data now
(generally not used)
RST, SYN, FIN:
connection estab
(setup, teardown
commands)
# bytes
rcvr willing
to accept
counting
by bytes
of data
(not segments!)
Internet
checksum
(as in UDP)
32. Transport Layer 3-59
TCP seq. numbers, ACKs
sequence numbers:
ī§byte stream ânumberâ
of first byte in
segmentâs data
acknowledgements:
ī§seq # of next byte
expected from other
side
ī§cumulative ACK
Q: how receiver handles
out-of-order segments
ī§A: TCP spec doesnât
say, - up to
implementor
source port # dest port #
sequence number
acknowledgement number
checksum
rwnd
urg pointer
incoming segment to sender
A
sent
ACKed
sent, not-
yet ACKed
(âin-
flightâ)
usable
but not
yet sent
not
usable
window size
N
sender sequence number space
source port # dest port #
sequence number
acknowledgement number
checksum
rwnd
urg pointer
outgoing segment from sender
33. Transport Layer 3-60
TCP seq. numbers, ACKs
User
types
âCâ
host ACKs
receipt
of echoed
âCâ
host ACKs
receipt of
âCâ, echoes
back âCâ
simple telnet scenario
Host BHost A
Seq=42, ACK=79, data = âCâ
Seq=79, ACK=43, data = âCâ
Seq=43, ACK=80
34. Transport Layer 3-62
RTT: gaia.cs.umass.edu to fantasia.eurecom.fr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RTT(milliseconds)
SampleRTT Estimated RTT
EstimatedRTT = (1- īĄ)*EstimatedRTT + īĄ*SampleRTT
īļ exponential weighted moving average
īļ influence of past sample decreases
exponentially fast
īļ typical value: īĄ = 0.125
TCP round trip time, timeout
RTT(milliseconds)
RTT: gaia.cs.umass.edu to fantasia.eurecom.fr
sampleRTT
EstimatedRTT
time (seconds)
35. Transport Layer 3-63
īļ timeout interval: EstimatedRTT plus âsafety
marginâ
ī§ large variation in EstimatedRTT -> larger safety margin
īļ estimate SampleRTT deviation from EstimatedRTT:DevRTT = (1-īĸ)*DevRTT +
īĸ*|SampleRTT-EstimatedRTT|
TCP round trip time, timeout
(typically, īĸ = 0.25)
TimeoutInterval = EstimatedRTT + 4*DevRTT
estimated RTT âsafety marginâ
36. Transport Layer 3-64
Chapter 3 outline
3.1 transport-layer
services
3.2 multiplexing and
demultiplexing
3.3 connectionless
transport: UDP
3.4 principles of
reliable data
transfer
3.5 connection-oriented
transport: TCP
ī§ segment structure
ī§ reliable data transfer
ī§ flow control
ī§ connection
management
3.6 principles of
congestion control
3.7 TCP congestion
control
37. Transport Layer 3-65
TCP reliable data transfer
īļ TCP creates rdt
service on top of IPâs
unreliable service
ī§ pipelined segments
ī§ cumulative acks
ī§ single retransmission
timer
īļ retransmissions
triggered by:
ī§ timeout events
ī§ duplicate acks
letâs initially consider
simplified TCP
sender:
ī§ ignore duplicate acks
ī§ ignore flow control,
congestion control
38. Transport Layer 3-66
TCP sender events:
data rcvd from app:
īļ create segment with
seq #
īļ seq # is byte-stream
number of first data
byte in segment
īļ start timer if not
already running
ī§ think of timer as for
oldest unacked
segment
ī§ expiration interval:
TimeOutInterval
timeout:
īļ retransmit segment
that caused timeout
īļ restart timer
ack rcvd:
īļ if ack acknowledges
previously unacked
segments
ī§ update what is known
to be ACKed
ī§ start timer if there are
still unacked
segments
39. Transport Layer 3-70
TCP ACK generation [RFC 1122, RFC 2581]
event at receiver
arrival of in-order segment with
expected seq #. All data up to
expected seq # already ACKed
arrival of in-order segment with
expected seq #. One other
segment has ACK pending
arrival of out-of-order segment
higher-than-expect seq. # .
Gap detected
arrival of segment that
partially or completely fills gap
TCP receiver action
delayed ACK. Wait up to 500ms
for next segment. If no next segment,
send ACK
immediately send single cumulative
ACK, ACKing both in-order segments
immediately send duplicate ACK,
indicating seq. # of next expected byte
immediate send ACK, provided that
segment starts at lower end of gap
40. Transport Layer 3-71
TCP fast retransmit
īļ time-out period
often relatively long:
ī§ long delay before
resending lost packet
īļ detect lost segments
via duplicate ACKs.
ī§ sender often sends
many segments
back-to-back
ī§ if segment is lost,
there will likely be
many duplicate
ACKs.
if sender receives 3
ACKs for same data
(âtriple duplicate
ACKsâ), resend
unacked segment
with smallest seq #
ī§ likely that unacked
segment lost, so
donât wait for timeout
TCP fast retransmit
(âtriple duplicate
ACKsâ),
41. Transport Layer 3-76
Chapter 3 outline
3.1 transport-layer
services
3.2 multiplexing and
demultiplexing
3.3 connectionless
transport: UDP
3.4 principles of
reliable data
transfer
3.5 connection-oriented
transport: TCP
ī§ segment structure
ī§ reliable data transfer
ī§ flow control
ī§ connection
management
3.6 principles of
congestion control
3.7 TCP congestion
control
42. Transport Layer 3-77
Connection Management
before exchanging data, sender/receiver
âhandshakeâ:
īļ agree to establish connection (each knowing the other
willing to establish connection)
īļ agree on connection parameters
connection state: ESTAB
connection variables:
seq # client-to-server
server-to-client
rcvBuffer size
at server,client
application
network
connection state: ESTAB
connection Variables:
seq # client-to-server
server-to-client
rcvBuffer size
at server,client
application
network
Socket clientSocket =
newSocket("hostname","port
number");
Socket connectionSocket =
welcomeSocket.accept();
43. Transport Layer 3-82
TCP: closing a connection
īļ client, server each close their side of
connection
ī§ send TCP segment with FIN bit = 1
īļ respond to received FIN with ACK
ī§ on receiving FIN, ACK can be combined with own
FIN
īļ simultaneous FIN exchanges can be handled
44. Transport Layer 3-84
Chapter 3 outline
3.1 transport-layer
services
3.2 multiplexing and
demultiplexing
3.3 connectionless
transport: UDP
3.4 principles of
reliable data
transfer
3.5 connection-oriented
transport: TCP
ī§ segment structure
ī§ reliable data transfer
ī§ flow control
ī§ connection
management
3.6 principles of
congestion control
3.7 TCP congestion
control
45. Transport Layer 3-85
congestion:
īļ informally: âtoo many sources sending too
much data too fast for network to handleâ
īļ different from flow control!
īļ manifestations:
ī§ lost packets (buffer overflow at routers)
ī§ long delays (queueing in router buffers)
īļ a top-10 problem!
Principles of congestion control
46. Transport Layer 3-92
R/2
lout
when sending at R/2,
some packets are
retransmissions
including duplicated
that are delivered!
âcostsâ of congestion:
īļ more work (retrans) for given âgoodputâ
īļ unneeded retransmissions: link carries multiple copies
of pkt
ī§ decreasing goodput
R/2lin
Causes/costs of congestion: scenario
2
Realistic: duplicates
īļ packets can be lost,
dropped at router due to
full buffers
īļ sender times out
prematurely, sending two
copies, both of which are
delivered
47. Transport Layer 3-95
Approaches towards congestion
control
two broad approaches towards congestion control:
end-end
congestion
control:
īļ no explicit feedback
from network
īļ congestion inferred
from end-system
observed loss,
delay
īļ approach taken by
TCP
network-assisted
congestion control:
īļ routers provide
feedback to end
systems
ī§single bit indicating
congestion (SNA,
DECbit, TCP/IP
ECN, ATM)
ī§explicit rate for
sender to send at
48. Transport Layer 3-96
Case study: ATM ABR congestion
control
ABR: available bit
rate:
īļ âelastic serviceâ
īļ if senderâs path
âunderloadedâ:
ī§ sender should use
available bandwidth
īļ if senderâs path
congested:
ī§ sender throttled to
minimum
guaranteed rate
RM (resource
management) cells:
īļ sent by sender,
interspersed with data cells
īļ bits in RM cell set by
switches (ânetwork-
assistedâ)
ī§ NI bit: no increase in
rate (mild congestion)
ī§ CI bit: congestion
indication
īļ RM cells returned to
sender by receiver, with
bits intact
49. Transport Layer 3-98
Chapter 3 outline
3.1 transport-layer
services
3.2 multiplexing and
demultiplexing
3.3 connectionless
transport: UDP
3.4 principles of
reliable data
transfer
3.5 connection-oriented
transport: TCP
ī§ segment structure
ī§ reliable data transfer
ī§ flow control
ī§ connection
management
3.6 principles of
congestion control
3.7 TCP congestion
control
50. Transport Layer 3-99
TCP congestion control: additive
increase multiplicative decrease
īļ approach: sender increases transmission rate
(window size), probing for usable bandwidth, until
loss occurs
ī§ additive increase: increase cwnd by 1 MSS
every RTT until loss detected
ī§ multiplicative decrease: cut cwnd in half after
loss
cwnd:TCPsender
congestionwindowsize
AIMD saw tooth
behavior: probing
for bandwidth
additively increase window size âĻ
âĻ. until loss occurs (then cut window in half)
time
51. Transport Layer 3-100
TCP Congestion Control:
details
īļ sender limits
transmission:
īļ cwnd is dynamic,
function of perceived
network congestion
TCP sending rate:
īļ roughly: send cwnd
bytes, wait RTT for
ACKS, then send
more bytes
last byte
ACKed sent, not-
yet ACKed
(âin-
flightâ)
last byte
sent
cwnd
LastByteSent-
LastByteAcked
< cwnd
sender sequence number space
rate ~~
cwnd
RTT
bytes/sec
52. Transport Layer 3-101
TCP Slow Start
īļ when connection
begins, increase rate
exponentially until first
loss event:
ī§ initially cwnd = 1 MSS
ī§ double cwnd every RTT
ī§ done by incrementing
cwnd for every ACK
received
īļ summary: initial rate is
slow but ramps up
exponentially fast
Host A
RTT
Host B
time
53. Transport Layer 3-102
TCP: detecting, reacting to
loss
īļ loss indicated by timeout:
ī§ cwnd set to 1 MSS;
ī§ window then grows exponentially (as in slow
start) to threshold, then grows linearly
īļ loss indicated by 3 duplicate ACKs: TCP
RENO
ī§ dup ACKs indicate network capable of
delivering some segments
ī§ cwnd is cut in half window then grows linearly
īļ TCP Tahoe always sets cwnd to 1 (timeout
or 3 duplicate acks)
54. Transport Layer 3-103
Q: when should the
exponential
increase switch
to linear?
A: when cwnd gets
to 1/2 of its
value before
timeout.
Implementation:
īļ variable ssthresh
īļ on loss event,
ssthresh is set to 1/2
of cwnd just before
loss event
TCP: switching from slow start to
CA
55. Transport Layer 3-104
Summary: TCP Congestion
Control
timeout
ssthresh = cwnd/2
cwnd = 1 MSS
dupACKcount = 0
retransmit missing segment
L
cwnd > ssthresh
congestion
avoidance
cwnd = cwnd + MSS (MSS/cwnd)
dupACKcount = 0
transmit new segment(s), as allowed
new ACK
.
dupACKcount++
duplicate ACK
fast
recovery
cwnd = cwnd + MSS
transmit new segment(s), as allowed
duplicate ACK
ssthresh= cwnd/2
cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeout
ssthresh = cwnd/2
cwnd = 1
dupACKcount = 0
retransmit missing segment
ssthresh= cwnd/2
cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3cwnd = ssthresh
dupACKcount = 0
New ACK
slow
start
timeout
ssthresh = cwnd/2
cwnd = 1 MSS
dupACKcount = 0
retransmit missing segment
cwnd = cwnd+MSS
dupACKcount = 0
transmit new segment(s), as allowed
new ACKdupACKcount++
duplicate ACK
L
cwnd = 1 MSS
ssthresh = 64 KB
dupACKcount = 0
New
ACK!
New
ACK!
New
ACK!
56. Transport Layer 3-105
TCP throughput
īļ avg. TCP thruput as function of window size,
RTT?
ī§ ignore slow start, assume always data to send
īļ W: window size (measured in bytes) where loss occurs
ī§ avg. window size (# in-flight bytes) is ž W
ī§ avg. thruput is 3/4W per RTT
W
W/2
avg TCP thruput =
3
4
W
RTT
bytes/sec
57. Transport Layer 3-106
TCP Futures: TCP over âlong, fat
pipesâ
īļ example: 1500 byte segments, 100ms RTT,
want 10 Gbps throughput
īļ requires W = 83,333 in-flight segments
īļ throughput in terms of segment loss
probability, L [Mathis 1997]:
â to achieve 10 Gbps throughput, need a loss rate
of L = 2Î10-10 â a very small loss rate!
īļ new versions of TCP for high-speed
TCP throughput =
1.22 . MSS
RTT L
58. Transport Layer 3-107
fairness goal: if K TCP sessions share same
bottleneck link of bandwidth R, each should
have average rate of R/K
TCP connection 1
bottleneck
router
capacity R
TCP Fairness
TCP connection 2
59. Transport Layer 3-108
Why is TCP fair?
two competing sessions:
īļ additive increase gives slope of 1, as throughout
increases
īļ multiplicative decrease decreases throughput
proportionallyR
R
equal bandwidth share
Connection 1 throughput
congestion avoidance: additive increase
loss: decrease window by factor of 2
congestion avoidance: additive increase
loss: decrease window by factor of 2
60. Transport Layer 3-109
Fairness (more)
Fairness and UDP
īļ multimedia apps
often do not use
TCP
ī§ do not want rate
throttled by
congestion control
īļ instead use UDP:
ī§ send audio/video at
constant rate,
tolerate packet loss
Fairness, parallel TCP
connections
īļ application can open
multiple parallel
connections between two
hosts
īļ web browsers do this
īļ e.g., link of rate R with 9
existing connections:
ī§ new app asks for 1 TCP, gets
rate R/10
ī§ new app asks for 11 TCPs, gets
R/2
61. Transport Layer 3-110
Chapter 3: summary
īļ principles behind
transport layer services:
ī§ multiplexing,
demultiplexing
ī§ reliable data transfer
ī§ flow control
ī§ congestion control
īļ instantiation,
implementation in the
Internet
ī§ UDP
ī§ TCP
next:
īļ leaving the
network âedgeâ
(application,
transport layers)
īļ into the network
âcoreâ