Transport Layer PDF

Chapter 3: Transport Layer
Our goals:
understand principles learn about transport
behind transport layer protocols in the
layer services: Internet:
multiplexing/ UDP: connectionless
demultiplexing transport
reliable data transfer TCP: connection-oriented
transport
flow control
TCP congestion control
congestion control
Transport Layer 3-1

Transport services and protocols
provide logical communication application
transport
between app processes network
data link
running on different hosts physical
network
data link
network physical
transport protocols run in
lo
data link
icg
physical
end systems
al
network
en
data link
d-
send side: breaks app physical network
en
data link
messages into segments,
d
physical
tr
an
passes to network layer network
sp
data link
or
physical
rcv side: reassembles
t
segments into messages, application
transport
passes to app layer network
data link
more than one transport physical
protocol available to apps

Internet: TCP and UDP
Transport Layer 3-2

Transport vs. network layer
network layer: logical Household analogy:
communication 12 kids sending letters
between hosts to 12 kids
transport layer: logical processes = kids
communication app messages = letters
between processes in envelopes
relies on, enhances,

hosts = houses
network layer services
transport protocol =
Ayşe and Bülent
network-layer protocol
= postal service
Transport Layer 3-3

Internet transport-layer protocols
reliable, in-order application
transport
delivery (TCP) network
data link network
physical
congestion control network
data link
physical
lo
data link
g
flow control
ic
physical
al
network
en
connection setup data link
d-
physical network
en
data link
unreliable, unordered
d
physical
tr
an
delivery: UDP network
sp
data link
or
physical
no-frills extension of
t

“best-effort” IP application
transport
network
services not available: data link
physical
delay guarantees
bandwidth guarantees
Transport Layer 3-4

Multiplexing/demultiplexing
Demultiplexing at rcv host: Multiplexing at send host:
gathering data from multiple
delivering received segments
sockets, enveloping data with
to correct socket
header (later used for
demultiplexing)
= socket = process
application P3 P1
P1 application P2 P4 application
transport transport transport
network network network
link link link
physical physical physical
host 2 host 3
host 1
Transport Layer 3-5
How demultiplexing works
host receives IP datagrams
each datagram has source 32 bits
IP address, destination IP
address source port # dest port #
each datagram carries 1

transport-layer segment other header fields
each segment has source,
destination port number
(recall: well-known port application
numbers for specific data
applications) (message)
host uses IP addresses & port
numbers to direct segment to
appropriate socket TCP/UDP segment format
Transport Layer 3-6

Connectionless demultiplexing
When host receives UDP
Create sockets with port
segment:
numbers:
DatagramSocket mySocket1 = new checks destination port
DatagramSocket(9111); number in segment
DatagramSocket mySocket2 = new directs UDP segment to
DatagramSocket(9222); socket with that port
number
UDP socket identified by
two-tuple: IP datagrams with
different source IP
(dest IP address, dest port number)
addresses and/or source
port numbers directed
to same socket
Transport Layer 3-7

Connection-oriented demux
TCP socket identified Server host may support
by 4-tuple: many simultaneous TCP
source IP address sockets:
source port number each socket identified by
dest IP address its own 4-tuple
dest port number Web servers have
recv host uses all four different sockets for
values to direct each connecting client
segment to appropriate non-persistent HTTP will
socket have different socket for
each request
Transport Layer 3-8

(cont)
P1 P4 P5 P6 P2 P1P3
SP: 5775
DP: 80
S-IP: B
D-IP:C
SP: 9157 SP: 9157

client DP: 80 server DP: 80 Client
IP: A IP: C S-IP: B
IP:B
S-IP: A
D-IP:C D-IP:C
Transport Layer 3-9

Threaded Web Server
P1 P4 P4
P5 P6 P2 P1P3
SP: 5775
DP: 80
S-IP: B
D-IP:C
SP: 9157 SP: 9157

client DP: 80 server DP: 80 Client
IP: A IP: C S-IP: B
IP:B
S-IP: A
D-IP:C D-IP:C
Transport Layer 3-10

UDP: User Datagram Protocol [RFC 768]
“no frills,” “bare bones”
Internet transport Why is there a UDP?
protocol
no connection
“best effort” service, UDP establishment (which can
segments may be: add delay)
lost simple: no connection state
delivered out of order at sender, receiver
to app small segment header
connectionless: no congestion control: UDP
no handshaking between can blast away as fast as
UDP sender, receiver desired
each UDP segment
handled independently
of others

UDP: more
often used for streaming
multimedia apps 32 bits
loss tolerant Length, in source port # dest port #

rate sensitive bytes of UDP length checksum
segment,
other UDP uses including
DNS header
SNMP
reliable transfer over UDP: Application
add reliability at data
application layer (message)
application-specific
error recovery!
UDP segment format

UDP checksum
Goal: detect “errors” (e.g., flipped bits) in transmitted
segment
Sender: Receiver:
treat segment contents compute checksum of
as sequence of 16-bit received segment
integers check if computed checksum
checksum: addition (1’s equals checksum field value:
complement sum) of NO - error detected
segment contents with YES - no error detected.
wraparound of carry out
bit
sender puts checksum
value into UDP checksum
field

Principles of Reliable data transfer
important in app., transport, link layers
top-10 list of important networking topics!
network
layer
characteristics of unreliable channel will determine

complexity of reliable data transfer protocol (rdt)
Reliable data transfer: getting started
rdt_send(): called from above, deliver_data(): called by
(e.g., by app.). Passed data to rdt to deliver data to upper
deliver to receiver upper layer
send receive
side side
udt_send(): called by rdt, rdt_rcv(): called when packet

to transfer packet over arrives on rcv-side of channel
unreliable channel to receiver

Reliable data transfer: getting started
We’ll:
incrementally develop sender, receiver sides of
reliable data transfer protocol (rdt)
consider only unidirectional data transfer
but control info will flow on both directions!
use finite state machines (FSM) to specify
sender, receiver
event causing state transition
actions taken on state transition
state: when in this
“state” next state state state
1 event
uniquely determined 2
by next event actions

Rdt1.0: reliable transfer over a reliable channel
underlying channel perfectly reliable

no bit errors
no loss of packets
separate FSMs for sender, receiver:
sender sends data into underlying channel
receiver read data from underlying channel
Wait for rdt_send(data) Wait for rdt_rcv(packet)

call from call from extract (packet,data)
above packet = make_pkt(data) below deliver_data(data)
udt_send(packet)
sender receiver

Rdt2.0: channel with bit errors
underlying channel may flip bits in packet
recall: checksum to detect bit errors
the question: how to recover from errors:
acknowledgements (ACKs): receiver explicitly tells sender
that pkt received OK
negative acknowledgements (NAKs): receiver explicitly
tells sender that pkt had errors
sender retransmits pkt on receipt of NAK
new mechanisms in rdt2.0 (beyond rdt1.0):
error detection
receiver feedback: control msgs (ACK,NAK) rcvr->sender

rdt2.0: FSM specification
rdt_send(data)
snkpkt = make_pkt(data, checksum) receiver
udt_send(sndpkt)
rdt_rcv(rcvpkt) &&
isNAK(rcvpkt)
Wait for Wait for rdt_rcv(rcvpkt) &&
call from ACK or udt_send(sndpkt) corrupt(rcvpkt)
above NAK
udt_send(NAK)
rdt_rcv(rcvpkt) && isACK(rcvpkt)

Wait for
Λ
call from
sender below
rdt_rcv(rcvpkt) &&
notcorrupt(rcvpkt)
extract(rcvpkt,data)
deliver_data(data)
udt_send(ACK)

rdt2.0: operation with no errors
rdt_send(data)
snkpkt = make_pkt(data, checksum)
udt_send(sndpkt)
rdt_rcv(rcvpkt) &&
isNAK(rcvpkt)
above NAK
udt_send(NAK)

Wait for
Λ call from
below
rdt_rcv(rcvpkt) &&
notcorrupt(rcvpkt)
deliver_data(data)
udt_send(ACK)

rdt2.0: error scenario
rdt_send(data)
snkpkt = make_pkt(data, checksum)
udt_send(sndpkt)
rdt_rcv(rcvpkt) &&
isNAK(rcvpkt)
above NAK
udt_send(NAK)

Wait for
Λ call from
below
rdt_rcv(rcvpkt) &&
notcorrupt(rcvpkt)
deliver_data(data)
udt_send(ACK)

rdt2.0 has a fatal flaw!
What happens if Handling duplicates:
ACK/NAK corrupted? sender adds sequence
sender doesn’t know what number to each pkt
happened at receiver! sender retransmits current
can’t just retransmit: pkt if ACK/NAK garbled
possible duplicate receiver discards (doesn’t
deliver up) duplicate pkt
What to do?
sender ACKs/NAKs
receiver’s ACK/NAK? What stop and wait
if sender ACK/NAK lost? Sender sends one packet,
then waits for receiver
retransmit, but this might
response
cause retransmission of
correctly received pkt!
rdt2.1: sender, handles garbled ACK/NAKs
rdt_send(data)
sndpkt = make_pkt(0, data, checksum)
udt_send(sndpkt) rdt_rcv(rcvpkt) &&
( corrupt(rcvpkt) ||
Wait for Wait for
ACK or
isNAK(rcvpkt) )
call 0 from
NAK 0 udt_send(sndpkt)
above
rdt_rcv(rcvpkt)
&& notcorrupt(rcvpkt) rdt_rcv(rcvpkt)
&& isACK(rcvpkt) && notcorrupt(rcvpkt)
&& isACK(rcvpkt)
Λ
Λ
Wait for Wait for
ACK or call 1 from
rdt_rcv(rcvpkt) && NAK 1 above
isNAK(rcvpkt) ) rdt_send(data)
udt_send(sndpkt) sndpkt = make_pkt(1, data, checksum)

udt_send(sndpkt)

rdt2.1: receiver, handles garbled ACK/NAKs
rdt_rcv(rcvpkt) && notcorrupt(rcvpkt)
&& has_seq0(rcvpkt)
deliver_data(data)
sndpkt = make_pkt(ACK, chksum)
udt_send(sndpkt)
rdt_rcv(rcvpkt) && (corrupt(rcvpkt) rdt_rcv(rcvpkt) && (corrupt(rcvpkt)
sndpkt = make_pkt(NAK, chksum) sndpkt = make_pkt(NAK, chksum)
udt_send(sndpkt) udt_send(sndpkt)
Wait for Wait for
rdt_rcv(rcvpkt) && 0 from 1 from rdt_rcv(rcvpkt) &&
not corrupt(rcvpkt) && below below not corrupt(rcvpkt) &&
has_seq1(rcvpkt) has_seq0(rcvpkt)
sndpkt = make_pkt(ACK, chksum) sndpkt = make_pkt(ACK, chksum)
udt_send(sndpkt) udt_send(sndpkt)
&& has_seq1(rcvpkt)
deliver_data(data)
sndpkt = make_pkt(ACK, chksum)
udt_send(sndpkt)

rdt2.1: discussion
Sender: Receiver:
seq # added to pkt must check if received
two seq. #’s (0,1) will packet is duplicate
suffice. Why? state indicates whether
0 or 1 is expected pkt
must check if received seq #
ACK/NAK corrupted
note: receiver can not
twice as many states know if its last
state must “remember” ACK/NAK received OK
whether “current” pkt
at sender
has 0 or 1 seq. #

rdt2.2: a NAK-free protocol
same functionality as rdt2.1, using ACKs only

instead of NAK, receiver sends ACK for last pkt
received OK
receiver must explicitly include seq # of pkt being ACKed
duplicate ACK at sender results in same action as
NAK: retransmit current pkt

rdt2.2: sender, receiver fragments
rdt_send(data)
sndpkt = make_pkt(0, data, checksum)
udt_send(sndpkt)
rdt_rcv(rcvpkt) &&
Wait for Wait for
ACK isACK(rcvpkt,1) )
call 0 from
above 0 udt_send(sndpkt)
sender FSM
fragment rdt_rcv(rcvpkt)
&& notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) && && isACK(rcvpkt,0)
(corrupt(rcvpkt) || Λ
has_seq1(rcvpkt)) Wait for receiver FSM
0 from
udt_send(sndpkt) below fragment
&& has_seq1(rcvpkt)
deliver_data(data)
sndpkt = make_pkt(ACK1, chksum)
udt_send(sndpkt) Transport Layer 3-27
rdt3.0: channels with errors and loss
New assumption: Approach: sender waits

underlying channel can “reasonable” amount of
also lose packets (data time for ACK
or ACKs) retransmits if no ACK
checksum, seq. #, ACKs, received in this time
retransmissions will be if pkt (or ACK) just delayed
of help, but not enough (not lost):
Q: how to deal with loss? retransmission will be
duplicate, but use of seq.
sender waits until
#’s already handles this
certain data or ACK
lost, then retransmits receiver must specify seq
# of pkt being ACKed
drawbacks?
requires countdown timer

rdt3.0 sender
rdt_send(data)
rdt_rcv(rcvpkt) &&
sndpkt = make_pkt(0, data, checksum) ( corrupt(rcvpkt) ||
udt_send(sndpkt) isACK(rcvpkt,1) )
rdt_rcv(rcvpkt) start_timer Λ
Λ Wait for Wait
for timeout
call 0from
ACK0 udt_send(sndpkt)
above
start_timer
rdt_rcv(rcvpkt)
&& notcorrupt(rcvpkt) rdt_rcv(rcvpkt)
&& isACK(rcvpkt,1) && notcorrupt(rcvpkt)
stop_timer && isACK(rcvpkt,0)
stop_timer
Wait Wait for
timeout for call 1 from
udt_send(sndpkt) ACK1 above
start_timer rdt_rcv(rcvpkt)
rdt_send(data) Λ
rdt_rcv(rcvpkt) &&
( corrupt(rcvpkt) || sndpkt = make_pkt(1, data, checksum)
isACK(rcvpkt,0) ) udt_send(sndpkt)
start_timer
Λ

rdt3.0 in action

rdt3.0 in action

Performance of rdt3.0
rdt3.0 works, but performance stinks

example: 1 Gbps link, 15 ms e-e prop. delay, 1KB packet:
Ttransmit = L (packet length in bits) 8kb/pkt

= = 8 microsec
R (transmission rate, bps) 10**9 b/sec
U L/R .008
= = = 0.00027
sender 30.008
RTT + L / R
U sender: utilization – fraction of time sender busy sending

1KB pkt every 30 msec -> 33kB/sec thruput over 1 Gbps link
network protocol limits use of physical resources!

rdt3.0: stop-and-wait operation
sender receiver
first packet bit transmitted, t = 0
last packet bit transmitted, t = L / R
first packet bit arrives

RTT last packet bit arrives, send
ACK
ACK arrives, send next

packet, t = RTT + L / R
U L/R .008
= = = 0.00027
sender 30.008
RTT + L / R

Pipelined protocols
Pipelining: sender allows multiple, “in-flight”, yet-to-
be-acknowledged pkts
range of sequence numbers must be increased
buffering at sender and/or receiver
Two generic forms of pipelined protocols: go-Back-N,

selective repeat
Pipelining: increased utilization
sender receiver
first packet bit transmitted, t = 0
last bit transmitted, t = L / R
first packet bit arrives

RTT last packet bit arrives, send ACK
last bit of 2nd packet arrives, send ACK
last bit of 3rd packet arrives, send ACK
ACK arrives, send next
packet, t = RTT + L / R
Increase utilization
by a factor of 3!
U 3*L/R .024
= = = 0.0008
sender 30.008
RTT + L / R

Utilization=N(L/R)/(RTT+L/R) if NL/R<RTT+L/R Utilization=1 if
and the sender pauses after it transmits a window NL/R>RTT+L/R and the
of packets until it receives first ACK sender does not pause

Go-Back-N
Sender:
k-bit seq # in pkt header
“window” of up to N, consecutive unack’ed pkts allowed
ACK(n): ACKs all pkts up to, including seq # n - “cumulative ACK”

may receive duplicate ACKs (see receiver)
timer for the entire window
timeout(n): retransmit pkt n and all higher seq # pkts in window

GBN: sender extended FSM
rdt_send(data)
if (nextseqnum < base+N) {
sndpkt[nextseqnum] = make_pkt(nextseqnum,data,chksum)
udt_send(sndpkt[nextseqnum])
if (base == nextseqnum)
start_timer
nextseqnum++
}
Λ else
refuse_data(data)
base=1
nextseqnum=1
timeout
start_timer
Wait
udt_send(sndpkt[base])
rdt_rcv(rcvpkt) udt_send(sndpkt[base+1])
&& corrupt(rcvpkt) …
udt_send(sndpkt[nextseqnum-1])
Λ
rdt_rcv(rcvpkt) &&
notcorrupt(rcvpkt)
base = getacknum(rcvpkt)+1
If (base == nextseqnum)
stop_timer
else
start_timer Transport Layer 3-38
GBN: receiver extended FSM
default
udt_send(sndpkt) rdt_rcv(rcvpkt)
&& notcurrupt(rcvpkt)
Λ && hasseqnum(rcvpkt,expectedseqnum)
expectedseqnum=1 Wait extract(rcvpkt,data)
sndpkt = deliver_data(data)
make_pkt(0,ACK,chksum) sndpkt = make_pkt(expectedseqnum,ACK,chksum)
udt_send(sndpkt)
expectedseqnum++
ACK-only: always send ACK for correctly-received pkt

with highest in-order seq #
may generate duplicate ACKs
need only remember expectedseqnum
out-of-order pkt:
discard (don’t buffer) -> no receiver buffering!
Re-ACK pkt with highest in-order seq #
GBN in
action

Selective Repeat
receiver individually acknowledges all correctly
received pkts
buffers pkts, as needed, for eventual in-order delivery
to upper layer
sender only resends pkts for which ACK not
received
sender timer for each unACKed pkt
sender window
N consecutive seq #’s
again limits seq #s of sent, unACKed pkts

Selective repeat: sender, receiver windows

Selective repeat
sender receiver
data from above : pkt n in [rcvbase, rcvbase+N-1]
if next available seq # in send ACK(n)
window, send pkt out-of-order: buffer
timeout(n): in-order: deliver (also
resend pkt n, restart timer deliver buffered, in-order
pkts), advance window to
ACK(n) in [sendbase,sendbase+N-1]: next not-yet-received pkt
mark pkt n as received
pkt n in [rcvbase-N,rcvbase-1]
if n smallest unACKed pkt,
ACK(n)
advance window base to next
unACKed seq # otherwise:
ignore

Selective repeat in action

Selective repeat:
dilemma
Example:
seq #’s: 0, 1, 2, 3
window size=3
receiver sees no
difference in two
scenarios!
incorrectly passes
duplicate data as new
in (a)
Q: what relationship
between seq # size
and window size?
Sequence Number vs. Window Size
Suppose we use k bits to represent SN
Question: What’s the minimum number of bits k
necessary for a window size of N?
Go-Back-N
Q: For a given expectedSN, what’s the largest possible value for
snd_base?
A: If all the last N ACKs sent by the receiver are received,
snd_base = expectedSN
snd_base=expectedSN expectedSN+N-1
sender
sender’s window
receiver
expectedSN
Go-Back-N
Q: For a given expectedSN, what’s the smallest possible value for
snd_base?
A: If all the last N ACKs sent by the receiver are not received,
snd_base = expectedSN-N
snd_base=expectedSN-N expectedSN-1
sender
sender’s window
receiver
expectedSN
Go-Back-N
All SNs in the interval [expectedSN-N,expectedSN+N-1] (an interval
of size 2N) can be received by the receiver. Since the receiver
accepts on the packet with SN=expectedSN, there should be no other
packet within this interval with SN=expectedSN. Therefore,
2k ≥ N+1
snd_base=expectedSN-N expectedSN+N-1
sender
receiver
expectedSN

Selective Repeat
Q: For a given rcv_base, what’s the largest possible value for
snd_base?
A: If all the last N ACKs sent by the receiver are received,
snd_base = rcv_base (same as go_back-N)
snd_base=rcv_base rcv_base+N-1
sender
sender’s window
receiver’s window
receiver
rcv_base rcv_base+N-1
Selective Repeat
Q: For a given rcv_base, what’s the smallest possible value for
snd_base?
A: If all the last N ACKs sent by the receiver are not received,
snd_base = rcv_base-N (same as Go-Back-N)
snd_base=rcv_base-N rcv_base-1
sender
sender’s window
receiver’s window
receiver
Selective Repeat
All SNs in the interval [rcv_base-N,rcv_base+N-1] (an interval of size
2N) can be received by the receiver. Since the receiver should be able
to distinguish between all packets in this interval and take
corresponding action, there should be no two packets within this
interval having the same SN. Therefore,
2k ≥ 2N
snd_base=rcv_base-N rcv_base+N-1
sender
receiver’s window
receiver
TCP: Overview RFCs: 793, 1122, 1323, 2018, 2581
point-to-point: full duplex data:

one sender, one receiver bi-directional data flow
reliable, in-order byte in same connection
stream: MSS: maximum segment
size
no “message boundaries”
connection-oriented:
pipelined:
handshaking (exchange
TCP congestion and flow of control msgs) init’s
control set window size sender, receiver state
send & receive buffers before data exchange
flow controlled:
sender will not
application application
socket
writes data reads data
socket
overwhelm receiver
door door
TCP TCP
send buffer receive buffer
segment

TCP segment structure
32 bits
URG: urgent data counting
(generally not used) source port # dest port #
by bytes
sequence number of data
ACK: ACK #
valid acknowledgement number (not segments!)
head not
PSH: push data now len used
UA P R S F Receive window
(generally not used) # bytes
checksum Urg data pnter
rcvr willing
RST, SYN, FIN: to accept
Options (variable length)
connection estab
(setup, teardown
commands)
application
Internet data
checksum (variable length)
(as in UDP)

TCP seq. #’s and ACKs
Seq. #’s:
byte stream “number” of first byte in segment’s data
ACKs:
seq # of next byte expected from other side
cumulative ACK
Q: how receiver handles out-of-order segments
A: TCP spec doesn’t say, - up to implementation
Widely used implementations of TCP buffer out-of-
order segments

TCP Round Trip Time and Timeout
Q: how to set TCP Q: how to estimate RTT?
timeout value? SampleRTT: measured time from
longer than RTT segment transmission until ACK
but RTT varies
receipt
ignore retransmissions
too short: premature
timeout SampleRTT will vary, want
unnecessary
estimated RTT “smoother”
retransmissions average several recent
too long: slow reaction

measurements, not just
to segment loss current SampleRTT

EstimatedRTT = (1- α)*EstimatedRTT + α*SampleRTT
Exponential weighted moving average

influence of past sample decreases exponentially fast
typical value: α = 0.125

Example RTT estimation:
RTT: gaia.cs.umass.edu to fantasia.eurecom.fr
350
300
250
RTT (milliseconds)
200
150
100
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
SampleRTT Estimated RTT

Setting the timeout
EstimatedRTT plus “safety margin”
large variation in EstimatedRTT -> larger safety margin
first estimate of how much SampleRTT deviates from
EstimatedRTT:
DevRTT = (1-β)*DevRTT +
β*|SampleRTT-EstimatedRTT|
(typically, β = 0.25)
Then set timeout interval:
TimeoutInterval = EstimatedRTT + 4*DevRTT

TCP reliable data transfer
TCP creates rdt Retransmissions are
service on top of IP’s triggered by:
unreliable service timeout events
Pipelined segments duplicate acks
Cumulative acks Initially consider
TCP uses single
simplified TCP sender:
ignore duplicate acks
retransmission timer;
ignore flow control,
however it just
congestion control
retransmits the first
segment in the window

TCP sender events:
data rcvd from app: timeout:
retransmit segment that
Create segment with
caused timeout (first
seq # segment in the window)
seq # is byte-stream restart timer
number of first data Ack rcvd:
byte in segment If acknowledges previously
start timer if not unacked segments
already running (think update what is known to
be acked
of timer as for oldest start timer if there are
unacked segment) outstanding segments
expiration interval:
TimeOutInterval

NextSeqNum = InitialSeqNum
SendBase = InitialSeqNum
loop (forever) { TCP

sender
switch(event)
event: data received from application above

create TCP segment with sequence number NextSeqNum (simplified)
if (timer currently not running)
start timer
pass segment to IP Comment:
NextSeqNum = NextSeqNum + length(data)
• SendBase-1: last
event: timer timeout cumulatively
retransmit not-yet-acknowledged segment with ack’ed byte
smallest sequence number Example:
start timer • SendBase-1 = 71;
y= 73, so the rcvr
event: ACK received, with ACK field value of y wants 73+ ;
if (y > SendBase) { y > SendBase, so
SendBase = y
that new data is
if (there are currently not-yet-acknowledged segments)
start timer
acked
}
} /* end of loop forever */

TCP: retransmission scenarios
Host A Host B Host A Host B
Seq=9 Seq=9
2 , 8 byt 2 , 8 byt
es da es da
ta
Seq=92 timeout
ta Seq=
1 00, 2
0 byt
es da
timeout
ta
=100
A CK 0
10
X C K=
= 120
loss A AC K
Seq=9 Seq=9
2
2 , 8 byt
es da Sendbase , 8 byt
es da
ta
ta = 100
Seq=92 timeout
SendBase
= 120 = 12
0
=100 A CK
A C K
SendBase
= 100 SendBase
= 120 premature timeout
time time
lost ACK scenario
TCP retransmission scenarios (more)
Host A Host B
Seq=9
2 , 8 byt
es da
ta
=100
timeout
Seq=1 A C K
0 0, 20
bytes
data
X
loss
SendBase C K =120
A
= 120
time
Cumulative ACK scenario

TCP ACK generation [RFC 1122, RFC 2581]
Event at Receiver TCP Receiver action

Arrival of in-order segment with Delayed ACK. Wait up to 500ms
expected seq #. All data up to for next segment. If no next segment,
expected seq # already ACKed send ACK
Arrival of in-order segment with Immediately send single cumulative

expected seq #. One other ACK, ACKing both in-order segments
segment has ACK pending
Arrival of out-of-order segment Immediately send duplicate ACK,

higher-than-expect seq. # . indicating seq. # of next expected byte
Gap detected
Arrival of segment that Immediate send ACK, provided that

partially or completely fills gap segment startsat lower end of gap

Fast Retransmit
Time-out period often If sender receives 3
relatively long: ACKs for the same
long delay before data, it supposes that
resending lost packet segment after ACKed
Detect lost segments data was lost:
via duplicate ACKs. fast retransmit: resend
Sender often sends segment before timer
many segments back-to- expires
back
If segment is lost,
there will likely be many
duplicate ACKs.

Fast Retransmit Host A Host B
Resend a segment
after 3 duplicate
ACKs since a seq # x1
duplicate ACK seq # x2
seq # x3
means that an out- seq # x4 X ACK x1
of sequence seq # x5
ACK x1
segment was ACK x1
ACK x1
received
triple
duplicate ACKs due duplicate
to packet ACKs resen
d seq X
reordering!
2
if window is small
don’t get duplicate timeout
ACKs!
time
Fast retransmit algorithm:
event: ACK received, with ACK field value of y

if (y > SendBase) {
SendBase = y
if (there are currently not-yet-acknowledged segments)
start timer
}
else {
increment count of dup ACKs received for y
if (count of dup ACKs received for y = 3) {
resend segment with sequence number y
}
a duplicate ACK for fast retransmit

already ACKed segment

TCP Flow Control
flow control
sender won’t overflow
receive side of TCP receiver’s buffer by
connection has a transmitting too much,
receive buffer: too fast
speed-matching
service: matching the
send rate to the
receiving app’s drain
rate
app process may be
slow at reading from
buffer
TCP Flow control: how it works
Rcvr advertises spare
room by including value
of RcvWin in segments
Sender limits unACKed
data to RcvWin
(Suppose TCP receiver guarantees receive
discards out-of-order buffer doesn’t overflow
segments)
spare room in buffer
= RcvWin
= RcvBuffer-[LastByteRcvd -
LastByteRead]

Sliding Window Flow Control
Example
Receiver
Buffer
Sender
0 4K
sends 2K
of data 2K SeqNo=0
2K
RcvWin =2048
Sender AckNo=2048
sends 2K
of data 2K SeqNo=2
048
Sender blocked
4K
N o=4 09 6 Rc vWin=0
Ack
3K
o=4 09 6 Rc vWin=1024
AckN

Principles of Congestion Control
Congestion:
informally: “too many sources sending too much
data too fast for network to handle”
different from flow control!
manifestations:
lost packets (buffer overflow at routers)
long delays (queueing in router buffers)
a top-10 problem!

Causes/costs of congestion: scenario 1
Host A λout
two senders, two λin : original data
receivers
one router,
Host B unlimited shared
output link buffers
infinite buffers
no retransmission
large delays
when congested
maximum
achievable
throughput
one router, finite buffers

sender retransmission of lost packet
Host A λin : original data λout
λ'in : original data, plus

retransmitted data
Host B finite shared output

link buffers

always: = λ
λ (goodput)
in out
“perfect” retransmission only when loss: λ > λout
in
retransmission of delayed (not lost) packet makes λ larger
in
(than perfect case) for same λout
R/2 R/2 R/2
λout R/3
λout
λout
R/4
R/2 R/2 R/2

λin λin λin
a. b. c.
“costs” of congestion:
more work (retrans) for given “goodput”
unneeded retransmissions: link carries multiple copies of pkt
four senders
Q: what happens as λ
multihop paths in
and λ increase ?
timeout/retransmit in
Host A λout
λin : original data
λ'in : original data, plus
retransmitted data
finite shared output

link buffers
Host B

H λ
o
o
s
u
t
A t
H
o
s
t
B
another “cost” of congestion:

when packet dropped, any “upstream transmission
capacity used for that packet was wasted!

Approaches towards congestion control
Two broad approaches towards congestion control:
End-end congestion Network-assisted

control: congestion control:
no explicit feedback from routers provide feedback
network to end systems
congestion inferred from single bit indicating
end-system observed loss, congestion (SNA,
delay DECbit, TCP/IP ECN,
approach taken by TCP ATM)
explicit rate sender
should send at

TCP Congestion Control
end-end control (no network How does sender
assistance) perceive congestion?
sender limits transmission: loss event = timeout or
LastByteSent-LastByteAcked 3 duplicate acks
≤ CongWin TCP sender reduces
CongWin is dynamic, function rate (CongWin) after
of perceived network loss event
congestion two modes of operation:
Slow Start (SS)
Congestion avoidance
(CA) or Additive
Increase Multiplicative
Decrease (AIMD)
TCP congestion control: bandwidth probing
“probing for bandwidth”: increase transmission rate
on receipt of ACK, until eventually loss occurs, then
decrease transmission rate
continue to increase on ACK, decrease on loss (since available
bandwidth is changing, depending on other connections in
network)
ACKs being received,
X loss, so decrease rate
so increase rate
X
X
X
sending rate
TCP’s
X “sawtooth”
behavior
time
Q: how fast to increase/decrease?

details to follow Transport Layer 3-79
TCP Congestion Control: details
sender limits rate by limiting number

of unACKed bytes “in pipeline”:
LastByteSent-LastByteAcked ≤ cwnd
cwnd: differs from rwnd (how, why?)
sender limited by min(cwnd,rwnd) cwnd

roughly, bytes
cwnd
rate = bytes/sec
RTT
RTT
cwnd is dynamic, function of
perceived network congestion ACK(s)

TCP Congestion Control: more details
segment loss event: ACK received: increase cwnd

reducing cwnd
Two modes of operation:
timeout: no response slowstart phase:
from receiver
• increase exponentially
cut cwnd to 1
fast (despite name)
3 duplicate ACKs: at at connection start,
least some segments or following timeout
getting through (recall congestion avoidance:
fast retransmit)
• increase linearly
cut cwnd in half, less
aggressively than on
timeout
TCP Slow Start Phase
when connection begins, cwnd = 1 Host A Host B
MSS
example: MSS = 500 bytes & one segm
ent
RTT = 200 msec
RTT
initial rate = 20 kbps
available bandwidth may be >>
two segm
en ts
MSS/RTT
desirable to quickly ramp up to
respectable rate four segm
ents
increase rate exponentially until
first loss event or when threshold
reached
double cwnd every RTT
done by incrementing cwnd by 1
time
for every ACK received

Slow Start Example
The congestion cwnd = 1 segment 1
window size grows ACK for segm

ent 1
very rapidly
cwnd = 2 segment 2
For every ACK, we
increase CongWin by
segment 3
1 irrespective of the ACK for segm

ents 2
number of segments cwnd = 3 ACK for segm
ents 3
ACK’ed cwnd = 4 segment 4
double CongWin segment 5
every RTT segment 6
initial rate is slow but segment 7
ramps up
exponentially fast ents 4
ACK for segm
TCP slows down the cwnd = 5 ACK for segm
ents 5
increase of CongWin cwnd = 6 ACK for segm

ents 6
when cwnd = 7 ACK for segm

ents 7
CongWin ≥ ssthresh cwnd = 8

TCP Congestion Avoidance Phase
when cwnd ≥ ssthresh AIMD
grow cwnd linearly
ACKs: increase cwnd
increase cwnd by 1
by 1 MSS per RTT:
MSS per RTT additive increase
approach possible
loss: cut cwnd in half
congestion slower (non-timeout-detected
than in slowstart loss ): multiplicative
implementation: cwnd decrease
= cwnd + MSS2/cwnd
for each ACK received AIMD: Additive Increase
Multiplicative Decrease

Congestion Avoidance
Congestion avoidance phase is started if CongWin has
reached the slow-start threshold value
If CongWin >= ssthresh then each time an ACK is

received, increment CongWin as follows:
• CongWin = CongWin + 1/CongWin (CongWin in
segments)
• In actual TCP implementation CongWin is in Bytes
CongWin = CongWin + MSS * (MSS/CongWin)
So CongWin is increased by one only if all CongWin
segments have been acknowledged.

Example Slow Start/
Congestion cwnd = 1
Avoidance
cwnd = 2
cwnd = 3
cwnd = 4
Assume that cwnd = 5
ssthresh = 8
cwnd = 6
cwnd = 7
cwnd = 8
14
12
Cwnd (in segments)
10 ssthresh
8
6
4 cwnd = 9
2
0
0
6
t=
t=
t=
t=
Roundtrip times
cwnd = 10
Slow Start / Congestion Avoidance
A typical plot of CongWin for a TCP connection
(MSS = 1500 bytes) with TCP Tahoe:
CA
ssthresh
SS

Responses to Congestion
TCP assumes there is congestion if it detects a packet
loss
A TCP sender can detect lost packets via loss events:
• Timeout of a retransmission timer
• Receipt of 3 duplicate ACKs (fast retransmit)
TCP interprets a Timeout as a binary congestion signal.
When a timeout occurs, the sender performs:
ssthresh is set to half the current size of the congestion
window:
ssthresh = CongWin / 2
CongWin is reset to one:
CongWin = 1
and slow-start is entered

Fast Recovery (differentiation
btwn two loss events)
After 3 dup ACKs (fast Philosophy:
Retransmit):
ssthresh = CongWin/2 • 3 dup ACKs indicates
network capable of
CongWin = CongWin/2
delivering some segments
window then grows
• timeout before 3 dup
linearly ACKs is “more alarming”
But after timeout event:
CongWin = 1 MSS;
window then grows
exponentially
to the threshold, then
grows linearly
TCP Congestion Control
Initially: Slow Start
CongWin = 1; (exponential
ssthresh = advertised window size; increase phase) is
New Ack received: continued until
if (CongWin < ssthresh) /* Slow Start*/ CongWin reaches
CongWin = CongWin + 1; half of the level
else /* Congestion Avoidance */ where the loss
CongWin = CongWin + 1/CongWin; event occurred
Timeout: last time.
ssthresh = CongWin/2; CongWin is
CongWin = 1; increased slowly
Fast Retransmission: after (linear
ssthresh = CongWin/2; increase in
CongWin = CongWin/2; Congestion
Avoidance phase).
3-90
Popular “flavors” of TCP
cwnd window size (in segments)
TCP Reno
ssthresh
ssthresh
TCP Tahoe
Transmission round

Summary: TCP Congestion Control
When CongWin is below Threshold, sender in slow-
start phase, window grows exponentially.
When CongWin is above Threshold, sender is in
congestion-avoidance phase, window grows linearly.
When a triple duplicate ACK occurs, Threshold set
to CongWin/2 and CongWin set to Threshold.
When timeout occurs, Threshold set to CongWin/2
and CongWin is set to 1 MSS.
The actual sender window size is determined based
on the congestion and flow control algorithms
SenderWin=min(RcvWin,CongWin)

TCP Congestion Control Summary
Event State TCP Sender Action Commentary
ACK receipt Slow Start CongWin = CongWin + MSS, Resulting in a doubling of
for previously (SS) If (CongWin ≥ Threshold) CongWin every RTT
unacked set state to “Congestion
data Avoidance”
ACK receipt Congestion CongWin = CongWin+MSS * Additive increase, resulting
for previously Avoidance (MSS/CongWin) in increase of CongWin by
unacked (CA) 1 MSS every RTT
data
Loss event SS or CA Threshold = CongWin/2, Fast recovery,
detected by CongWin = Threshold, implementing multiplicative
triple Set state to “Congestion decrease. CongWin will not
duplicate Avoidance” drop below 1 MSS.
ACK
Timeout SS or CA Threshold = CongWin/2, Enter slow start
CongWin = 1 MSS,
Set state to “Slow Start”
Duplicate SS or CA Increment duplicate ACK count CongWin and Threshold not
ACK for segment being acked changed

TCP throughput
Q: what’s average throughout of TCP as
function of window size, RTT?
ignoring slow start
let W be window size when loss occurs.
when window is W, throughput is W/RTT
just after loss, window drops to W/2,
throughput to W/2RTT.
average throughout: .75 W/RTT

TCP Futures: TCP over “long, fat pipes”
example: 1500 byte segments, 100ms RTT, want 10

Gbps throughput
requires window size W = 83,333 in-flight
segments
throughput in terms of loss rate:
1.22 ⋅ MSS
RTT L
L = 2·10-10 Wow
new versions of TCP for high-speed

TCP Fairness
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
TCP
router
connection 2
capacity R

Why is TCP fair?
Two competing sessions:
Additive increase gives slope of 1, as throughout increases
multiplicative decrease decreases throughput proportionally
R equal bandwidth share

Connection 2 throughput
loss: decrease window by factor of 2

congestion avoidance: additive increase
loss: decrease window by factor of 2
congestion avoidance: additive increase
Connection 1 throughput R

Fairness (more)
Fairness and UDP Fairness and parallel TCP
Multimedia apps often
connections
do not use TCP nothing prevents app from
do not want rate opening parallel cnctions
throttled by congestion between 2 hosts.
control Web browsers do this
Instead use UDP: Example: link of rate R
pump audio/video at supporting 9 cnctions;
constant rate, tolerate
new app asks for 1 TCP, gets
packet loss
rate R/10
Research area: TCP new app asks for 11 TCPs,
friendly gets R/2 !

TCP Connection Management
Recall: TCP sender, receiver Three way handshake:
establish “connection”
before exchanging data Step 1: client host sends TCP
segments SYN segment to server
initialize TCP variables: specifies initial seq #
seq. #s no data
buffers, flow control Step 2: server host receives
info (e.g. RcvWindow) SYN, replies with SYNACK
client: connection initiator segment
Socket clientSocket = new
Socket("hostname","port
server allocates buffers
number"); specifies server initial
seq. #
server: contacted by client
Socket connectionSocket = Step 3: client receives SYNACK,
welcomeSocket.accept(); replies with ACK segment,
which may contain data

TCP Connection Management (cont.)
Closing a connection: client server
client closes socket: close

FIN
clientSocket.close();
Step 1: client end system AC K

sends TCP FIN control close
segment to server FIN
Step 2: server receives
timed wait
A CK
FIN, replies with ACK.
Closes connection, sends
FIN.
closed

TCP Connection Management (cont.)
Step 3: client receives FIN, client server

replies with ACK. closing
FIN
Enters “timed wait” -
will respond with ACK
to received FINs AC K
closing
Step 4: server, receives FIN
ACK. Connection closed.
timed wait
A CK
closed
closed

TCP Connection Management (cont)
TCP server
lifecycle
TCP client
lifecycle

Tuning TCP/IP Parameters
TCP/IP parameters
A set of default values may not be optimal for all applications.
The network administrator may wish to turn on or off some
TCP/IP functions for performance or security considerations.
Many Unix and Linux systems provide some flexibility in
tuning the TCP/IP kernel.
/sbin/sysctl is used to configure the Linux kernel
parameters at runtime.
Default kernel configuration file is /sbin/sysctl.conf.
Frequently used sysctl options:
• sysctl –a or sysctl –A: list all current values.
• sysctl –p file_name: load the sysctl setting from a configuration
file.
• sysctl –w variable=value: change the value of the parameter
SomeTCP Parameters in Linux Kernel
tcp_syn_retries
Number of SYN packets the kernel will send before giving up on
the new connection.
tcp_synack_retries
number of SYN+ACK packets sent before the kernel gives up on
the connection.
tcp_window_scaling
Maximum window size of 65535 bytes not enough for for really
fast networks. The window scaling options allows for almost
gigabyte windows, which is good for connections with large
delay-bandwidth product.
tcp_max_syn_backlog
Maximal number of remembered connection requests, which
still did not receive an acknowledgment from connecting client.
tcp_fin_timeout
How many seconds to wait for a final FIN packet before
the socket is closed; required to prevent denial-of-service
(DoS) attacks. Default value is 60 seconds.
SomeTCP Parameters in Linux Kernel
tcp_rmem
This is a vector of 3 integers: [min, default, max].
These parameters are used by TCP to dynamically adjust
receive buffer sizes.
min - minimum size of the receive buffer used by each
TCP socket. The default value is 4K.
default - the default size of the receive buffer for a
TCP socket. The default value is 87380 bytes, and is
lowered to 43689 in low memory systems. If larger
receive buffer sizes are desired, this value should be
increased.
max - the maximum size of the receive buffer used by
each TCP socket. The default value of 87380*2 bytes is
lowered to 87380 in low memory systems.
tcp_smem
Send buffer parameters [min, default, max] similar to
tcp_rmem. Transport Layer 3-105

Transport Layer PDF

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Transport Layer PDF

Enviado por

Direitos autorais:

Formatos disponíveis

Chapter 3: Transport Layer

Transport Layer 3-1

protocol available to apps

Transport Layer 3-2

Transport Layer 3-3

Transport Layer 3-4

transport transport transport

network network network

link link link

physical physical physical

 each datagram carries 1

Transport Layer 3-6

Transport Layer 3-7

Transport Layer 3-8

SP: 9157 SP: 9157

Transport Layer 3-9

SP: 9157 SP: 9157

Transport Layer 3-10

Transport Layer 3-11

 loss tolerant Length, in source port # dest port #

Transport Layer 3-12

Transport Layer 3-13

 characteristics of unreliable channel will determine

udt_send(): called by rdt, rdt_rcv(): called when packet

Transport Layer 3-15

Transport Layer 3-16

 underlying channel perfectly reliable

Wait for rdt_send(data) Wait for rdt_rcv(packet)

Transport Layer 3-17

Transport Layer 3-18

rdt_rcv(rcvpkt) && isACK(rcvpkt)

Transport Layer 3-19

rdt_rcv(rcvpkt) && isACK(rcvpkt)

Transport Layer 3-20

rdt_rcv(rcvpkt) && isACK(rcvpkt)

Transport Layer 3-21

udt_send(sndpkt) sndpkt = make_pkt(1, data, checksum)

Transport Layer 3-23

Transport Layer 3-24

Transport Layer 3-25

 same functionality as rdt2.1, using ACKs only

Transport Layer 3-26

New assumption: Approach: sender waits

Transport Layer 3-28

Transport Layer 3-29

Transport Layer 3-30

Transport Layer 3-31

 rdt3.0 works, but performance stinks

Ttransmit = L (packet length in bits) 8kb/pkt

 U sender: utilization – fraction of time sender busy sending

Transport Layer 3-32

first packet bit arrives

ACK arrives, send next

Transport Layer 3-33

 Two generic forms of pipelined protocols: go-Back-N,

first packet bit arrives

Transport Layer 3-35

Transport Layer 3-36

 ACK(n): ACKs all pkts up to, including seq # n - “cumulative ACK”

Transport Layer 3-37

ACK-only: always send ACK for correctly-received pkt

Transport Layer 3-40

Transport Layer 3-41

each datagram carries 1

loss tolerant Length, in source port # dest port #

characteristics of unreliable channel will determine

underlying channel perfectly reliable

same functionality as rdt2.1, using ACKs only

rdt3.0 works, but performance stinks

U sender: utilization – fraction of time sender busy sending

Two generic forms of pipelined protocols: go-Back-N,

ACK(n): ACKs all pkts up to, including seq # n - “cumulative ACK”

point-to-point: full duplex data:

too long: slow reaction

Exponential weighted moving average

one router, finite buffers

Q: how fast to increase/decrease?

sender limits rate by limiting number

sender limited by min(cwnd,rwnd) cwnd

If CongWin >= ssthresh then each time an ACK is

example: 1500 byte segments, 100ms RTT, want 10