Documente Academic
Documente Profesional
Documente Cultură
LAN
Requirements
EEE
802.11b
Subnets
and
Subnet
Masks
Issues
to
Address
Internet
Mail
Architecture
•
Throughput
•
Extension
of
802.11
DSSS
(direct
sequence
spread
spectrum)
scheme
•
Allows
arbitrary
complexity
of
internetworked
LANs
within
organiza;on
•Ordered
delivery:
With
an
unreliable
network
service
it
is
possible
that
segments
may
•
Currently
defined
in
RFC
5598
•
Fundamental
level
consists
of:
•
Number
of
nodes
–
Data
rates
of
5.5
and
11
Mbps
•
Insulate
overall
internet
from
growth
of
network
numbers
and
rou;ng
complexity
arrive
out
of
order.
Solu;on
to
this
problem
is
to
number
segments
sequen;ally.
TCP
uses
–
A
user
world,
in
the
form
of
Message
User
Agents
(MUA)
•
Connec;on
to
backbone
LAN
•
Chipping
rate
11
MHz
•
Site
looks
to
rest
of
internet
like
single
network
•
Each
LAN
assigned
subnet
number
•
Host
scheme
where
each
data
octet
is
implicitly
numbered
–
The
transfer
world,
in
the
form
of
the
Message
Handling
Service
(MHS),
which
is
composed
of
Message
Transfer
Agents
(MTA
•
Service
area
–
Same
as
original
DSSS
scheme
–
Complementary
Code
Keying
por;on
of
address
par;;oned
into
subnet
number
and
host
•Retransmission
strategy:
Events
necessita;ng
retransmission:
Sending
transport
does
not
•
BaJery
power
consump;on
(CCK)
modula;on
gives
higher
data
rate
with
same
–
Packet
Binary
Convolu;onal
Coding
number
•
Local
routers
route
within
subneJed
network
•
Subnet
mask
indicates
which
bits
know
transmission
was
unsuccessful
(Segment
may
be
damaged
in
transit
but
s;ll
arrives
at
Administra;ve
Management
Domain
(ADMD)
•
Transmission
robustness
and
security
(PBCC)
for
future
higher
rate
us
are
subnet
number
and
which
are
its
des;na;on
OR
Segment
fails
to
arrive).
Sending
transport
does
not
know
transmission
Ø
An
Internet
e-‐mail
provider
•
Collocated
network
opera;on
host
number
was
unsuccessful.
Receiver
acknowledges
successful
receipt
by
returning
a
segment
•
Each
ADMD
can
have
different
opera;ng
policies
and
trust-‐based
decision
making
•
License-‐free
opera;on
IEEE
802.11g
containing
an
acknowledgment
number.
Examples:
•
Handoff/roaming
•
Higher-‐speed
extension
to
802.11b
•
Operates
in
2.4GHz
band
•
Compa;ble
with
802.11b
Internet
Control
Message
Protocol
(ICMP)
•Duplicate
detec7on:
Receiver
must
be
able
to
recognize
duplicates
•
Segment
sequence
•
A
department
that
operates
a
local
mail
relay
(MTA)
•
An
IT
department
that
operates
an
enterprise
mail
relay
•
An
ISP
that
operates
a
public
shared
e-‐
•
Dynamic
configura;on
devices
•
Combines
physical
layer
encoding
techniques
used
in
802.11
and
802.11b
to
provide
•
RFC
792
numbers
help
•
Complica;ons
arise
if:
A
duplicate
is
received
prior
to
the
close
of
the
mail
service
service
at
a
variety
of
data
rates
•
Provides
a
means
for
transferring
messages
from
routers
and
other
hosts
to
a
host
connec;on
(Sender
must
not
get
confused
if
it
receives
mul;ple
acknowledgments
to
the
Wi-‐Fi
Alliance
–
ERP-‐OFDM
for
6,
9,
12,
18,
24,
36,
48,
54Mbps
rates
–
ERP-‐ •
Provides
feedback
about
problems
same
segment
AND
Sequence
number
space
must
be
long
enough)
OR
IF
A
duplicate
is
Internet
Mail
Standards
•
There
is
always
a
concern
whether
products
from
different
vendors
will
successfully
interoperate
PBCC
for
22
and
33Mbps
rates
•
Datagram
cannot
reach
its
des;na;on
•
Router
does
not
have
buffer
capacity
to
forward
•
received
aher
the
close
of
the
connec;on
•
Allows
an
e-‐mail
client
to
download
an
e-‐mail
from
an
e-‐mail
server
•
Wireless
Ethernet
Compa;bility
Alliance
(WECA)
Router
can
send
traffic
on
a
shorter
route
•
Connects
via
TCP/IP
to
the
server
–
Industry
consor;um
formed
in
1999
IEEE
802.11n
•
Encapsulated
in
IP
datagram
•Flow
control:
Future
acknowledgments
will
resynchronize
the
protocol
if
an
ACK/CREDIT
Post
Office
Protocol
(POP3)
•
Renamed
the
Wi-‐Fi
(Wireless
Fidelity)
Alliance
•
Has
enhancements
in
three
general
areas:
–
Hence
not
reliable
segment
is
lost.
•
If
no
new
acknowledgments
are
forthcoming
the
sender
;mes
out
and
•
Provides
stronger
authen;ca;on
than
POP3
and
provides
other
func;ons
not
supported
by
POP3
–
Created
a
test
suite
to
cer;fy
interoperability
for
802.11
product
–
Mul;ple-‐input-‐mul;ple-‐output
(MIMO)
antenna
architecture
retransmits
a
data
segment
which
triggers
a
new
acknowledgment.
•
S;ll
possible
for
Internet
Mail
Access
Protocol
(IMAP)
•
Most
important
enhancement
deadlock
to
occur
•
Protocol
used
for
transfer
of
mail
from
a
user
agent
to
an
MTA
and
from
one
MTA
to
another
–
Radio
transmission
scheme
Simple
Mail
Transfer
Protocol
(SMTP)
Distribu7on
of
Messages
Within
a
DS
•
Increased
capacity
Common
ICMP
Messages
•Connec7on
establishment:•
Must
take
into
account
the
unreliability
of
a
network
service
•
•
Supplements
SMTP
and
allows
the
encapsula;on
of
mul;media
messages
inside
of
a
standard
SMTP
message
Integra;on
service
•
Des;na;on
unreachable
•
Time
exceeded
•
Parameter
problem
•
Source
quench
•
Redirect
Calls
for
the
exchange
of
SYNs
(two
way
handshake)
–
Could
result
in:
•
Duplicate
SYNs
•
–
MAC
enhancements
Mul;purpose
Internet
Mail
Extensions
(MIME)
•Enables
transfer
of
data
between
a
sta;on
on
an
IEEE
802.11
LAN
and
a
sta;on
on
an
integrated
IEEE
802.x
•
Echo
and
echo
reply
•
Timestamp
and
;mestamp
reply
•
Address
mask
request
and
reply
•
MLost
AN
significant
change
is
to
aggregate
mul;ple
MAC
frames
into
a
single
block
for
Duplicate
data
segments
•Takes
care
of
any
address
transla;on
and
media
conversion
logic
required
for
the
exchange
of
data
transmission
SMTP
Address
Resolu7on
Protocol
(ARP)
•Connec7on
termina7on:
Two-‐way
handshake
was
found
to
be
inadequate
for
an
unreliable
•
Defined
in
RFC
821
Distribu;on
service
•
Need
MAC
address
to
send
to
LAN
host
network
service.
•
Out
of
order
segments
could
cause
the
FIN
segment
to
arrive
before
the
•
The
standard
protocol
for
transferring
mail
between
hosts
in
the
TCP/IP
suite
•Primary
service
used
by
sta;ons
to
exchange
MAC
frames
when
frame
must
traverse
the
DS
to
get
from
a
sta;on
in
one
BSS
to
a
sta;on
in
another
Manual
Included
in
network
address
Use
central
directory
Use
address
resolu;on
protocol
last
data
segment
-‐
To
avoid
this
problem
the
next
sequence
number
aher
the
last
octet
of
•
Is
not
concerned
with
the
format
or
content
of
messages
themselves
Connec7onless
Opera7on
•ARP
(RFC
826)
provides
dynamic
IP
to
Ethernet
address
mapping
BSS
•
Internetworking
involves
connec;onless
opera;on
at
the
level
of
the
Internet
Protocol
(IP)
data
can
be
assigned
to
FIN
-‐
Each
side
must
explicitly
acknowledge
the
FIN
of
the
other
–
Excep;ons:
•If
sta;ons
are
in
the
same
BSS,
distribu;on
service
logically
goes
through
the
single
AP
of
that
BSS
Source
broadcasts
ARP
request
-‐-‐
Des;na;on
replies
with
ARP
response
using
an
ACK
with
the
sequence
number
of
the
FIN
to
be
acknowledged
•
SMTP
standardizes
the
message
character
set
as
7-‐bit
ASCII
•
IP
Ini;ally
developed
for
the
DARPA
internet
project
•
Protocol
is
needed
to
access
a
par;cular
network
•
SMTP
adds
log
informa;on
to
the
start
of
the
delivered
message
that
indicates
the
path
the
message
took
IP
Next
Genera7on
Failure
Recovery:
•
When
the
system
that
the
transport
en;ty
is
running
on
fails
and
•
Connec;onless
internet
facility
is
flexible
•
IP
provides
a
connec;onless
service
between
Requirements
for
new
types
of
service
subsequently
restarts,
the
state
informa;on
of
all
ac;ve
connec;ons
is
lost
–
Affected
Limita;ons
of
SMTP/822
end
systems.
Advantages:
•
Is
flexible
•
Can
be
made
robust
•
Does
not
impose
unnecessary
•
Address
configura;on
rou;ng
flexibility
•
Traffic
support
Associa7on-‐Related
Services
overhead
connec;ons
become
half
open
because
the
side
that
did
not
fail
does
not
realize
•
DS
requires
informa;on
about
sta;ons
within
the
ESS
that
is
provided
by
the
associa;on-‐related
services
the
problem
(S;ll
ac;ve
side
of
a
half-‐open
connec;on
can
close
the
connec;on
using
a
SMTP
cannot
transmit
executable
files
or
other
binary
objects
•
Sta;on
must
be
associated
before
DS
can
deliver
data
to
or
accept
data
from
it
IP
Design
Issues
keepalive
;mer)
SMTP
cannot
transmit
text
data
that
includes
na;onal
language
characters
•
3
mobility
transi;on
types:
•
In
the
event
that
a
transport
en;ty
fails
and
quickly
restarts,
half-‐open
connec;ons
can
be
SMTP
servers
may
reject
mail
messages
over
a
certain
size
•
Rou;ng
•
Datagram
life;me
•
Fragmenta;on
and
reassembly
•
Error
control
•
Flow
control
ddress
space
exhaus;on:
1)
No
transi;on
sta;onary
or
in
single
BSS
•
Two
level
addressing
(network
and
host)
wastes
space
terminated
more
quickly
by
the
the
use
of
the
RST
segment
–
Failed
side
returns
an
RST
i
to
SMTP
gateways
that
translate
between
the
character
codes
ASCII
and
EBCDIC
do
not
use
a
consistent
set
of
mappings,
resul;ng
in
transla;on
problems
2)
BSS
transi;on
between
BSS
in
same
ESS
•
Network
addresses
used
even
if
not
connected
every
segment
i
that
it
receives
–
RST
i
must
be
checked
for
validity
on
the
other
side
(If
valid
SMTP
gateways
to
X.400
electronic
mail
networks
cannot
handle
nontextual
data
included
in
X.400
message
3)
ESS
transi;on
between
BSS
in
different
ESS
•
Growth
of
networks
and
the
Internet
an
abnormal
termina;on
occurs)
Some
SMTP
implementa;ons
do
not
adhere
completely
to
the
SMTP
standards
defined
in
RFC
82
Rou7ng
•
DS
needs
to
know
the
iden;ty
of
the
AP
to
which
the
message
should
be
delivered
(ES
/
routers
maintain
rou;ng
tables)
•
Extended
use
of
TCP/IP
•
Single
address
per
host
•
There
is
s;ll
the
chance
that
some
user
data
will
be
lost
or
duplicated
–
Sta;ons
must
maintain
associa;on
with
MIME
•
Rou;ng
table
indicates
next
router
to
which
datagram
is
sent
AP
within
current
BSS
•
Can
be
sta;c
or
dynamic
IPv6
RFCs
TCP
Services
•
Extension
to
the
RFC
822
framework
that
is
intended
to
address
some
of
the
problems
and
limita;ons
of
the
use
of
SMTP
and
RFC
822
for
electronic
mail
3
services
relate
to
this
requirement:
(Source
rou;ng)
•
RFC
1752
-‐
Recommenda;ons
for
the
IP
Next
Genera;on
Protocol
–
Requirements
–
PDU
•
TCP
labels
data
as:
•
Data
stream
Push
•
Includes
the
following
elements:
•
Associa;on
-‐
establishes
ini;al
associa;on
between
sta;on
and
AP
formats
–
Addressing,
rou;ng
security
issues
•
Ordinarily,
TCP
decides
when
sufficient
data
have
accumulated
to
form
a
segment
for
–
Five
new
message
header
fields
are
defined,
which
may
be
included
in
an
•
Source
specifies
route
to
be
followed
•
Reassocia;on
–
enables
an
established
associa;on
to
be
transferred
from
one
AP
to
another
•
Can
be
useful
for
security
and
priority
•
RFC
2460
-‐
overall
specifica;on
transmission.
RFC
822
header
•
Disassocia;on
–
a
no;fica;on
from
either
a
sta;on
or
an
AP
that
an
exis;ng
associa;on
is
terminated
•
RFC
4291
-‐
addressing
structure
•
The
TCP
user
can
require
TCP
to
transmit
all
outstanding
data
up
to
and
including
that
–
A
number
of
content
formats
are
defined,
thus
standardizing
representa;ons
that
support
mul;media
electronic
mail
(Route
recording)
•
Each
router
appends
its
internet
address
to
a
list
of
addresses
in
the
datagram
labeled
with
a
push
flag.
–
Transfer
encodings
are
defined
that
enable
the
conversion
of
any
content
format
into
a
form
that
is
protected
from
altera;on
by
the
mail
syste
•
Useful
for
tes;ng
and
debugging
purposes
IPv6
Enhancements
•
Urgent
data
signaling
•
Expanded
128
bit
address
space
•
Improved
op;on
mechanism
•
This
provides
a
means
of
informing
the
des;na;on
TCP
user
that
significant
or
"urgent"
MIME
Header
Fields
Medium
Access
Control
•
The
rou;ng
table
may
be
sta;c
or
dynamic.
MAC
layer
covers
three
func;onal
areas:
•
A
sta;c
table,
however,
could
contain
alternate
routes
if
a
–
Most
not
examined
by
intermediate
routes
data
is
in
the
upcoming
data
stream.
It
is
up
to
the
des;na;on
user
to
determine
appropriate
•
MIME-‐version
•
Content-‐type
1)
Reliable
data
delivery
•
Dynamic
address
assignment
•
Increased
addressing
flexibility
ac;on.
–
Describes
the
data
contained
in
the
body
with
sufficient
detail
that
the
receiving
user
agent
can
pick
an
appropriate
agent
or
mechanism
to
present
the
par;cular
router
is
unavailable.
2)
Access
control
•
A
dynamic
table
is
more
flexible
in
responding
to
both
error
–
Anycast
and
mul;cast
•
Defined
in
terms
of
primi;ves
and
parameters
data
to
the
use
3)
Security
and
conges;on
condi;ons.
•
Support
for
resource
alloca;on
•
Content-‐transfer-‐encoding
•
For
example,
when
a
router
goes
–
Labeled
packet
flows
TCP
Mechanisms
–
Indicates
the
type
of
transforma;on
that
has
been
used
to
represent
the
body
of
the
message
in
a
way
that
is
acceptable
for
mail
transport
Reliable
Data
Delivery
down,
all
of
its
neighbors
will
send
out
a
status
report,
allowing
other
routers
and
sta;ons
to
•
Can
be
grouped
into:
•
Content-‐ID
•
Content-‐descrip;on
Can
be
dealt
with
at
a
higher
layer
update
their
rou;ng
tables.
Connec;on
establishment
•
More
efficient
to
deal
with
errors
at
MAC
level
•
Conges;on
control
is
par;cularly
•
Always
uses
a
three-‐way
handshake
MIME
Content
Types
IPv6
Flow
Label
•
802.11
includes
frame
exchange
protocol
important
because
of
the
mismatch
in
capacity
between
local
and
wide
area
networks
•
Related
sequence
of
packets
•
Special
handling
•
Iden;fied
by
source
and
des;na;on
•
Connec;on
is
determined
by
host
and
port
POP
Data
transfer
•
Internet
standard
defined
in
RFC
1939
–
Sta;on
receiving
frame
returns
acknowledgment
(ACK)
frame
address
plus
flow
label
•
Router
treats
flow
as
sharing
aJributes
•
May
treat
flows
differently
–
Exchange
treated
as
atomic
unit,
not
to
be
interrupted
by
a
transmission
from
any
other
sta;on
•
Alterna;ve
to
including
all
informa;on
in
every
header
•
Have
requirements
on
flow
label
•
Viewed
logically
as
consis;ng
of
a
stream
of
octets
•
Supports
the
basic
func;ons
of
download
and
delete
for
e-‐mail
retrieval
–
If
no
ACK
within
short
period
of
;me,
retransmit
processing
•
Flow
control
is
exercised
using
credit
alloca;on
•
States:
Datagram
Life7me
Connec;on
termina;on
Authen;ca;on
state
•
If
dynamic
or
alternate
rou;ng
is
used
the
poten;al
exists
for
a
datagram
to
loop
•
802.11
physical
and
MAC
layers
unreliable
IPv6
Addresses
•
Each
TCP
user
must
issue
a
CLOSE
primi;ve
Client
must
authen;cate
itself
to
the
user
indefinitely
•
An
abrupt
termina;on
occurs
if
the
user
issues
an
ABORT
primi;ve
Ohen
done
with
a
user
ID/password
combina;on
–
Noise,
interference,
and
other
propaga;on
effects
result
in
loss
of
frames
–
Consumes
resources
•
128
bits
long
•
Assigned
to
interface
•
Single
interface
may
have
mul;ple
unicast
addresses
–
Even
with
error-‐correc;on
codes,
frames
may
not
successfully
be
received
•
Unicast
-‐
single
interface
address
Transac;on
state
–
Transport
protocol
may
need
upper
bound
on
life;me
of
a
•
Anycast
-‐
one
of
a
set
of
interface
addresses
TCP
Implementa7on
Policy
Op7ons
Once
the
server
successfully
authen;cates
the
client,
the
client
can
access
the
mailbox
to
retrieve
and
delete
messages
datagram
•
Implementa;on
opportuni;es:
Update
state
•
Can
mark
datagram
with
life;me
•
Mul;cast
-‐
all
of
a
set
of
interfaces
Send
policy
Deliver
policy
Accept
policy
Retransmit
policy
Acknowledge
policy
During
this
state,
the
server
enacts
all
of
the
changes
requested
by
the
client’s
commands
and
then
closes
the
connec;on
Four
Frame
Exchange
•
When
life;me
expires,
datagram
is
discarded
•
RTS
alerts
all
sta;ons
within
range
of
source
that
exchange
is
under
way
Fragmenta7on
Header
•
Fragmenta;on
only
allowed
at
source
•
No
fragmenta;on
at
intermediate
routers
•
Node
Send
Policy
IMAP
•
CTS
alerts
all
sta;ons
within
range
of
des;na;on
Fragmenta;on
and
Re-‐assembly
•
Other
sta;ons
don’t
transmit
to
avoid
collision
must
perform
path
discovery
to
find
•
In
the
absence
of
both
pushed
data
and
a
closed
transmission
window
a
sending
TCP
en;ty
•
Defined
by
RFC
3501
•
Protocol
exchanges
data
between
two
en;;es
is
free
to
transmit
data
at
its
own
convenience
•
Servers
store
messages
for
mul;ple
users
to
be
retrieved
upon
client
requests
•
RTS/CTS
exchange
is
a
required
func;on
of
MAC
but
may
be
disabled
•
Lower-‐level
protocols
may
need
to
break
data
up
into
smaller
blocks,
called
fragmenta;on
smallest
MTU
of
intermediate
networks
•
Set
source
fragments
to
match
MTU
•
Otherwise
limit
to
1280
octets
•
TCP
may
construct
a
segment
for
each
batch
of
data
provided
or
it
may
wait
un;l
a
certain
•
Features:
•
Source
issues
a
Request
to
Send
(RTS)
frame
à
Des;na;on
responds
with
Clear
to
Send
(CTS)
-‐-‐>
Aher
•
receiving
Reasons
CfTS,
or
fsragmenta;on:
ource
transmits
data
-‐-‐>
amount
of
data
accumulates
before
construc;ng
and
sending
a
segment
Des;na;on
responds
with
ACK
–
Clients
can
have
mul;ple
remote
mailboxes
from
which
messages
can
be
–
Network
only
accepts
blocks
of
a
certain
size
–
More
efficient
•
Infrequent
and
large
transmissions
have
low
overhead
in
terms
of
segment
genera;on
and
retrieved
error
control
and
smaller
retransmission
units
–
Fairer
access
to
shared
facili;es
–
Smaller
Distributed
Coordina7on
Func7on
(DCF)
buffers
processing
–
Clients
can
also
specify
criteria
for
downloading
messages,
such
as
not
transferring
large
messages
over
slow
links
Rou7ng
Header
•
If
transmissions
are
frequent
and
small,
the
system
is
providing
quick
response
–
IMAP
always
keeps
messages
on
the
server
and
replicates
copies
to
the
clients
DCF
sublayer
uses
CSMA
algorithm
•
Disadvantages:
•
Contains
a
list
of
one
or
more
intermediate
nodes
to
be
visited
on
the
way
to
a
packet’s
–
IMAP4
allows
clients
to
make
changes
both
when
connected
and
when
disconnected
•
Does
not
include
a
collision
detec;on
func;on
because
it
is
not
prac;cal
on
a
wireless
network
–
Smaller
buffers
–
More
interrupts
and
processing
;me
des;na;on
Deliver
Policy
•
Includes
a
set
of
delays
that
amounts
as
a
priority
scheme
Header
includes:
•
Next
header
•
Header
extension
length
•
Rou;ng
type
•
Segments
leh
•
In
the
absence
of
a
Push,
a
receiving
TCP
en;ty
is
free
to
deliver
data
to
the
user
at
its
own
(If
sta;on
has
frame
to
send
it
listens
to
medium.
If
medium
is
idle,
sta;on
may
transmit.
Else
waits
un;l
current
transmission
is
complete.
Domain
Name
System
(DNS)
convenience
•
A
directory
lookup
service
that
provides
a
mapping
between
the
name
of
a
host
on
the
Internet
and
its
numerical
address
IP
Fragmenta7on
Des7na7on
Op7ons
Header
•
May
deliver
as
each
in-‐order
segment
is
received,
or
may
buffer
data
before
delivery
•
IP
re-‐assembles
at
des;na;on
only
•
Uses
fields
in
header
•Carries
op;onal
informa;on
for
des;na;on
node
•
If
deliveries
are
infrequent
and
large,
the
user
is
not
receiving
data
as
promptly
as
may
be
Internet
Directory
Service
(DNS)
Priority
IFS
Values
•Format
same
as
hop-‐by-‐hop
header
desirable
•
Essen;al
to
the
func;oning
of
the
Internet
•
Defined
in
RFCs
1034
and
1035
•
Four
elements
comprise
the
DNS:
•
SIFS
(short
IFS)
Error
and
Flow
Control
•
If
deliveries
are
frequent
and
small,
there
may
be
unnecessary
processing,
as
well
as
Domain
name
space
For
all
immediate
response
ac;ons
•
Error
control
Connec7on-‐Oriented
Transport
Mechanisms
•
PIFS
opera;ng
system
interrupts
DNS
database
–
Discarded
datagram
iden;fica;on
is
needed
•
Two
basic
types
of
transport
service:
Name
servers
(point
coordina;on
func;on
IFS)
–
Reasons
for
discarded
datagrams
include:
Connec;on-‐oriented
Accept
Policy
Resolvers
Used
by
the
centralized
controller
in
PCF
scheme
when
issuing
polls
•
Life;me
expira;on
•
Conges;on
•
Establishment,
maintenance
and
termina;on
of
a
logical
connec;on
between
TS
users
•
If
segments
arrive
out
of
order
the
receiving
TCP
en;ty
has
two
op;ons:
•
DIFS
•
FCS
error
•
Has
a
wide
variety
of
applica;ons
•
Most
common
•
Implies
service
is
reliable
In
order
Domain
(distributed
coordina;on
func;on
IFS)
Connec;onless
or
datagram
service
Used
as
minimum
delay
for
asynchronous
frames
contending
for
access
•
Accepts
only
segments
that
arrive
in
order;
any
segment
that
•
Refers
to
a
group
of
hosts
that
are
under
the
administra;ve
control
of
a
single
en;ty
•
Flow
control
arrives
out
of
order
is
discarded
•
Organized
hierarchically,
so
that
a
given
domain
may
consist
of
a
number
of
subordinate
domains
–
Allows
routers
to
limit
the
rate
they
receive
data
Reliable
Sequencing
Network
Service
•
Makes
for
simple
implementa;on
but
places
a
burden
on
the
networking
facility
•
Names
are
assigned
and
reflect
the
hierarchical
organiza;on
•
Any
sta;on
using
SIFS
to
determine
transmission
opportunity
has
the
highest
priority
–
Send
flow
control
packets
reques;ng
reduced
data
flow
•
Issues:
•
If
a
single
segment
is
lost
in
transit,
then
all
subsequent
segments
must
be
retransmiJed
•
Used
in
the
following
circumstances:
Addressing
Mul;plexing
Flow
control
Connec;on
establishment/termina;on
In
window
DNS
Database
–
Acknowledgment
(ACK)
Internet
Protocol
(IP)
v4
•
Sta;on
responds
with
an
ACK
frame
aher
wai;ng
only
for
a
SIFS
gap
•
Provides
for
efficient
collision
recovery
•
Accepts
all
segments
that
are
within
the
receive
window
•
Based
on
a
hierarchical
database
containing
resource
records
(RRs)
that
include
the
name,
IP
address,
and
other
informa;on
about
hosts
•
Defined
in
RFC
791
•
Part
of
TCP/IP
suite
•
Two
parts
Addressing
•
Requires
a
more
complex
acceptance
test
and
a
more
sophis;cated
data
storage
scheme
•
Key
features:
–
Clear
to
Send
(CTS)
•
Specifica;on
of
interface
with
a
higher
layer
•
Transport
protocol
must
be
able
to
derive
the
following
informa;on
from
the
TS
user
•
Sta;on
ensures
data
frame
gets
through
by
issuing
RTS
–
Variable-‐depth
hierarchy
for
names
–
Distributed
database
–
Distribu;on
controlled
by
the
database
•
Specifica;on
of
actual
protocol
format
and
mechanisms
address:
Retransmit
Policy
–
Poll
response
–
User
iden;fica;on
–
Transport
en;ty
iden;fica;on
–
Host
address
–
Network
number
•
Retransmission
strategies:
Name
Resolu;on
IP
Services
First
only
•
Each
query
begins
at
a
name
resolver
located
in
the
user
host
system
•
Primi;ves
Mul7plexing
•
Maintain
one
retransmission
;mer
for
en;re
queue
•
If
the
resolver
does
not
have
the
requested
name
in
its
cache,
it
sends
a
DNS
query
to
the
local
DNS
server
Four
Frame
Exchange
–
Specifies
func;ons
to
be
performed
•
Mul;ple
users
employ
the
same
transport
protocol
and
are
dis;nguished
by
port
numbers
•
Frame
Control:
Indicates
the
type
of
frame
(control,
management,
or
data)
•
Dura;on/Connec;on
ID:
T
he
;me
the
channel
will
be
allocated
•
If
an
ack
is
received,
remove
the
appropriate
segment(s)
from
the
queue
and
reset
the
•
Resolvers
use
UDP
for
single
queries
and
TCP
for
group
queries
–
Form
off
or
primi;ve
implementa;on
dependent
or
service
access
points
;mer.
Recursive
technique
successful
transmission
of
a
MAC
frame
or
connec;on,
iden;fier.
•
Addresses:
The
number
and
meaning
•
of
Ptarameters
he
48-‐bit
address
fields
depend
on
Upward
mul;plexing
context.
MAC
address
of
transmiJer,
receiver,
and
AP.
•
Sequence
Control:
Contains
a
4-‐bit
fragment
number
used
for
fragmenta;on,
and
a
12-‐bit
•
If
the
;mer
expires,
retransmit
the
segment
at
the
front
of
the
queue
and
reset
the
;mer.
•
Query
another
name
server
for
the
desired
result
and
then
send
the
result
back
–
Used
to
pass
data
and
control
informa;on
•
Mul;plexing
of
mul;ple
connec;ons
on
a
single
lower-‐level
connec;on
sequence
number
used
to
number
frames
•
QoS
Control:
IEEE
802.11
QoS
facility
•
High
Throughput
Control:
control
bits
related
to
the
•
Efficient
in
terms
of
traffic
generated
•
Can
have
considerable
delays
•
Used
for
requests
sent
by
a
name
resolver
–
o pera;on
of
transmission
of
data
unit
–
Deliver-‐no;fy
user
of
Send-‐request
Downward
mul;plexing
Itera;ve
technique
802.11n,
802.11ac,
and
802.11ad.
•
Frame
Body:
LLC
PDU
or
MAC
control
informa;on.
•
Frame
Check
Sequence:
arrival
of
Ad
32-‐bit
ata
CRC
unit
•
Spliqng
of
a
single
connec;on
•
Return
the
address
of
the
next
server
to
whom
the
request
should
be
sent
then
send
out
a
new
DNS
request
to
that
server
among
mul;ple
lower-‐level
connec;ons
IP
Parameters
Batch
•
Maintain
one
retransmission
;mer
for
en;re
queue
HTTP
•
Source
and
des;na;on
addresses
•
Protocol
•
Type
of
Service
•
Iden;fica;on
•
Don’t
Flow
Control
Control
Frames
•
Transac;on-‐oriented
client/server
protocol
•
Most
typical
use
is
between
a
Web
browser
and
a
Web
server
•
Makes
use
of
TCP
to
provide
reliability
•
Is
a
•
Request
AP
transmit
buffered
frame
when
in
power-‐saving
fragment
indicator
•
Time
to
live
•
Data
length
•
Op;on
data
•
User
data
•
Complex
at
the
transport
layer:
•
if
an
ack
is
received,
remove
the
appropriate
segment(s)
from
the
queue
and
reset
the
;mer.
stateless
protocol
•
Each
transac;on
is
treated
independently
•
Flexible
in
the
formats
that
it
can
handle
mode
–
Considerable
delay
in
the
communica;on
of
flow
control
Power
Save-‐Poll
(PS-‐Poll)
informa;on
•
If
the
;mer
expires,
retransmit
all
segments
in
the
queue
and
reset
the
;mer.
Proxy
•
First
frame
in
four-‐way
frame
exchange
IP
Op7ons
–
Amount
of
the
transmission
delay
may
be
highly
variable,
•
Reduces
the
likelihood
of
long
delays
•
May
result
in
unnecessary
retransmissions
•
A
forwarding
agent,
receiving
a
request
for
a
URL
object,
modifying
the
request,
and
forwarding
the
request
toward
the
server
iden;fied
in
the
URL
•
Security
–
Allows
a
security
label
to
be
aJached
to
a
datagram.
making
it
difficult
to
effec;vely
use
a
;meout
mechanism
for
retransmission
of
lost
data
Request
to
Send
(RTS)
•
Acts
as
a
server
in
interac;ng
with
a
client
and
as
a
client
in
interac;ng
with
a
server
•
Second
frame
in
four-‐way
exchange
•
Source
rou;ng
–
A
sequenced
list
of
router
addresses
that
specifies
the
route
to
be
•
Reasons
for
control:
Individual
•
Maintain
one
;mer
for
each
segment
in
the
queue
•
More
complex
implementa;on
•
Scenarios
that
call
for
the
use
of
a
proxy:
followed.
Rou;ng
may
be
strict
(only
iden;fied
routers
may
be
visited)
or
loose
(other
User
of
the
receiving
transport
en;ty
cannot
keep
up
with
the
flow
Clear
to
Send
(CTS)
–
Security
intermediary
•
Acknowledges
correct
receipt
intermediate
routers
may
be
visited).
Receiving
transport
en;ty
itself
cannot
keep
up
with
the
flow
of
segments
•
Client
and
server
may
be
separated
by
a
security
intermediary
such
as
a
firewall,
with
the
proxy
on
the
client
side
of
the
firewall
Acknowledgment
(ACK)
•
Route
recording
–
A
field
is
allocated
to
record
the
sequence
of
routers
visited
by
the
Acknowledge
Policy
•
Timing
of
acknowledgment:
–
Different
versions
of
HTTP
datagram.
Alterna7ves
to
Flow
Control
Requirements
(recieving
en7ry
can...)
•
Announces
end
of
conten;on-‐free
period
part
of
PCF
•
If
the
client
and
server
are
running
different
versions
of
HTTP,
the
proxy
can
implement
both
versions
and
perform
the
required
mapping
Conten;on-‐Free
(CF)-‐end
•
Stream
iden;fica;on
–
Names
reserved
resources
used
for
stream
service.
This
service
•
Do
nothing
-‐
Segments
that
overflow
the
buffer
are
discarded
-‐
Sending
transport
en;ty
will
Immediate
•
Immediately
transmit
an
empty
segment
containing
the
appropriate
acknowledgement
provides
special
handling
for
vola;le
periodic
traffic
(e.g.,
voice).
retransmit
•
Acknowledges
CF-‐end
to
end
conten;on-‐free
period
and
release
sta;ons
from
associated
restric;ons
Gateway
•
Timestamping
–
The
source
IP
en;ty
and
some
or
all
intermediate
routers
add
a
;mestamp
•
Refuse
to
accept
further
segments
from
the
network
service
-‐
Relies
on
network
service
to
number
•
Simple
and
keeps
the
remote
TCP
fully
informed
•
A
server
that
appears
to
the
client
as
if
it
were
an
origin
server
(precision
to
milliseconds)
to
the
data
unit
as
it
goes
by.
do
the
work.
•
Acts
on
behalf
of
other
servers
that
may
not
be
able
to
communicate
directly
with
a
client
•
Use
a
fixed
sliding
window
protocol
-‐
With
a
reliable
network
service
this
works
quite
well
•
Limits
unnecessary
retransmissions
Data
Frames
IP
Address
Classes
•
Results
in
extra
segment
transmissions
•
Scenarios
in
which
gateways
can
be
used:
•
Use
a
credit
scheme
-‐
A
more
effec;ve
scheme
to
use
with
an
unreliable
network
service
–
Security
intermediary
•
Eight
data
frame
subtypes
Class
A
•
Can
cause
a
further
load
on
the
network
•
The
client
and
server
may
be
separated
by
a
security
intermediary
such
as
a
firewall,
with
the
gateway
on
the
server
side
of
the
firewall.
–
Organized
in
two
groups
•Start
with
binary
0
Connec7on
Establishment
and
Termina7on
Cumulu;ve
•
First
four
carry
upper-‐level
data
•Network
addresses
with
a
first
octet
of
0
(binary
0000000)
and
127
(binary
01111111)
are
•
Wait
for
an
outbound
segment
with
data
on
which
to
piggyback
the
acknowledgement
•
In
this
case
the
client
must
authen;cate
itself
to
the
gateway,
which
can
then
pass
the
request
on
to
the
server.
•
Serves
three
main
purposes:–
Allows
each
end
to
assure
that
the
other
exists
–
Allows
–
Non-‐HTTP
server
•
Remaining
do
not
carry
any
user
reserved
•
Typically
used
exchange
or
nego;a;on
of
op;onal
parameters
–
Triggers
alloca;on
of
transport
en;ty
•
The
client
makes
an
HTTP
request
to
a
•
gateway
server.
The
gateway
server
then
contacts
the
relevant
FTP
or
Gopher
•
server
to
obtain
the
desired
data
•126
poten;al
Class
A
network
numbers
resources.
(Is
by
mutual
agreement)
•
Requires
more
processing
at
the
receiving
end
and
complicates
the
task
of
es;ma;ng
•
Data
•Range
1
to
126
round-‐trip
;me
result.
–
Simplest
data
frame,
conten;on
or
conten;on-‐free
uClass
se
B
Unreliable
Network
Service
•
Data
+
CF-‐Ack
•Start
with
binary
10
User
Datagram
Protocol
(UDP)
Tunnel
Examples:
-‐
Internetwork
using
IP
-‐
IEEE
802.3
LAN
using
the
unacknowledged
connec;onless
•
Performs
no
opera;ons
on
HTTP
requests
and
responses
•
Is
simply
a
relay
point
between
two
TCP
connec;ons
•
Used
when
there
must
be
an
–
Carries
data
and
acknowledges
previously
received
•Range
data
d 128
uring
to
191(binary
10000000
to
10111111)
•
Transport-‐level
protocol
that
is
commonly
used
as
part
of
the
TCP/IP
protocol
suite
LLC
service
(Segments
are
occasionally
lost
and
may
arrive
out
of
sequence
due
to
variable
intermediary
system
between
client
and
server
but
it
is
conten;on-‐free
period
•Second
octet
also
included
in
network
address
transit
delays
•
RFC
768
•
Data
+
CF-‐Poll
•2^14
=
16,384
Class
B
addresses
•
Provides
a
connec;onless
service
for
applica;on-‐level
procedures
not
necessary
for
that
system
to
understand
the
contents
of
messages
arrive
out
of
sequence
due
to
variable
transit
delays
–
Used
by
point
coordinator
to
deliver
data
and
request
Class
send
C
•
Unreliable
service;
delivery
and
duplicate
protec;on
are
not
guaranteed
•
Data
+
CF-‐Ack
+
CF-‐Poll
•Start
with
binary
110
•
Reduces
overhead
and
may
be
adequate
in
many
cases
Cache
–
Combines
Data
+
CF-‐Ack
and
Data
+
CF-‐Poll
•Range
192
to
223
•
A
facility
that
may
store
previous
requests
and
responses
for
handling
new
requests
•Second
and
third
octet
also
part
of
network
address
–
If
a
new
request
arrives
that
is
the
same
as
a
stored
request
then
the
cache
can
supply
the
stored
response
rather
than
accessing
the
Management
Frames
•2^21
=
2,097,152
addresses
resource
indicated
in
the
URL
Used
to
manage
communica;ons
between
sta;ons
and
APs
•Nearly
all
allocated
(See
IPv6)
•
Can
operate
on
a
client
or
server
or
on
an
intermediate
system
other
than
a
tunnel
Management
of
associa;ons
•
Not
all
transac;ons
can
be
cached
•
Request,
response,
reassocia;on,
dissocia;on,
and
authen;ca;on
General
Header
Fields
•
Can
be
used
in
both
request
and
response
messages
•
Are
applicable
in
both
types
of
messages
and
contain
informa;on
that
does
not
directly
apply
to
the
en;ty
being
transferred
•
Fields
are:
IO
queues
Window
Management
Decision
Time
and
Place
Approaches
to
Rou;ng
Mul;cas;ng:
The
act
of
sending
a
packet
from
a
source
to
the
members
of
a
mul;cast
group
•
On
each
port,
packets
arrive
and
depart.
•
The
size
of
TCP’s
send
window
can
have
a
cri;cal
effect
on
whether
TCP
can
be
used
Decision
;me
•
Internet
rou;ng
protocols
employ
one
of
three
approaches
to
gathering
and
using
rou;ng
informa;on:
•
Mul;cast
addresses
•
Consider
that
there
are
two
buffers
at
each
port,
one
to
accept
arriving
packets,
and
one
to
efficiently
without
causing
conges;on
•
Packet
or
virtual
circuit
basis
•
Fixed
or
dynamically
changing
Distance-‐vector
rou;ng
–
Addresses
that
refer
to
a
group
of
hosts
on
one
or
more
networks
hold
packets
wai;ng
to
depart
•
Two
techniques
found
in
virtually
all
modern
implementa;on
of
TCP
are:
Decision
place
Link-‐state
rou;ng
Has
a
number
of
prac;cal
applica;ons
–
There
might
be
two
fixed-‐size
buffers
–
Slow
start
•
Distributed
-‐
made
by
each
node
Path-‐vector
rou;ng
Mul;media
“broadcast”
Teleconferencing
Database
Distributed
compu;ng
Real
;me
associated
with
each
port,
–
Dynamic
window
sizing
on
conges;on
•
More
complex,
but
more
robust
workgroups
–
or
there
might
be
a
pool
of
memory
available
for
all
buffering
ac;vi;es.
•
Centralized
–
made
by
a
designated
node
•
Source
–
made
by
source
sta;on
Distance-‐Vector
Rou;ng
•
As
packets
arrive,
they
are
stored
in
the
input
buffer
of
the
corresponding
port.
Slow
Start
•
Requires
that
each
node
exchange
informa;on
with
its
neighboring
nodes
LAN
Mul;cast
•
Node
examines
each
packet,
makes
a
rou;ng
decision,
and
then
moves
the
packet
to
the
•
Gradually
expand
the
window
un;l
acknowledgments
are
received
•
TCP
transmission
is
Network
Informa;on
Source
and
Update
Timing
–
Two
nodes
are
said
to
be
neighbors
if
they
are
both
directly
connected
to
the
same
network
•
LAN
mul;cast
is
easy
appropriate
output
buffer.
constrained
by
•
Rou;ng
decisions
usually
based
on
knowledge
of
network,
traffic
load,
and
link
cost
•
Used
in
the
first-‐genera;on
rou;ng
algorithm
for
ARPANET
–
Send
to
IEEE
802
mul;cast
MAC
address
–
Those
in
mul;cast
group
will
accept
it
–
Only
awnd
=
MIN[credit,
cwnd]
–
Distributed
rou;ng
•
Each
node
maintains
a
vector
of
link
costs
for
each
directly
aJached
network
and
distance
and
next-‐ single
copy
of
packet
is
needed
If
node
6
restrains
the
flow
of
packets
from
node
5,
this
causes
the
output
buffer
in
node
5
for
where
•
Using
local
knowledge,
informa;on
from
adjacent
nodes,
informa;on
from
all
nodes
on
a
hop
vectors
for
each
des;na;on
•
A
transmission
from
any
one
sta;on
is
received
by
all
other
sta;ons
on
LAN
the
port
to
node
6
to
fill
up.
•
awnd
=
allowed
window,
in
segment
poten;al
route
•
Rou;ng
Informa;on
Protocol
(RIP)
uses
this
approach
Conges;on
at
one
point
in
the
network
can
quickly
propagate
throughout
a
region
or
the
–
This
is
the
number
of
segments
that
TCP
is
currently
allowed
to
send
without
receiving
–
Central
rou;ng
Mul;cas;ng
Strategies
en;re
network.
further
acknowledgments
•
Collect
informa;on
from
all
nodes
Link-‐State
Rou;ng
•
If
server
does
not
know
members
of
group
While
flow
control
is
indeed
a
powerful
tool,
we
need
to
use
it
in
such
a
way
as
to
manage
the
•
cwnd
=
conges;on
window,
in
segments
Issue
of
update
;ming
•
Designed
to
overcome
the
drawbacks
of
distance-‐vector
rou;ng
•
When
a
router
is
ini;alized,
it
Broadcast
packet
to
each
network
traffic
on
the
en;re
network.
–
A
window
used
by
TCP
during
startup
and
to
reduce
flow
during
periods
of
conges;on
•
Depends
on
rou;ng
strategy
•
Fixed
-‐
never
updated
•
Adap;ve
-‐
regular
updates
determines
the
link
cost
on
each
of
its
network
•
To
each
network
with
members
in
mul;cast
group
•
credit
=
the
amount
of
unused
credit
granted
in
the
most
recent
acknowledgment,
in
interfaces
Could
send
mul;ple
unicast
packets
Backpressure
segments
Rou;ng
Strategies
-‐
Fixed
Rou;ng
•
The
router
then
adver;ses
this
set
of
link
costs
to
all
other
routers
in
the
internet
topology,
not
just
•
Spanning
tree
•
Replicated
by
routers
at
branch
points
•
If
node
becomes
congested
it
can
slow
down
or
stop
flow
of
packets
from
other
nodes
–
When
an
acknowledgment
is
received,
this
value
is
calculated
as
window/segment_size,
•
Use
a
single
permanent
route
for
each
source
to
des;na;on
pair
of
nodes
neighboring
routers
True
mul;cast
•
Can
be
exerted
on
the
basis
of
links
or
logical
connec;ons
where
•
Determined
using
a
least
cost
algorithm
•
Route
is
fixed
•
From
then
on,
the
router
monitors
its
link
costs
•
Flow
restric;on
propagates
backward
to
sources,
which
are
restricted
in
the
flow
of
new
window
is
a
field
in
the
incoming
TCP
segment
(the
amount
of
data
the
peer
TCP
en;ty
is
–
Un;l
a
change
in
network
topology
–
Based
on
expected
traffic
or
capacity
•
Whenever
there
is
a
significant
change
the
router
again
adver;ses
its
set
of
link
costs
to
all
other
Requirements
for
Mul;cas;ng
packets
into
the
network
willing
to
accept)
•
Advantage
is
simplicity
•
Disadvantage
is
lack
of
flexibility
routers
in
the
configura;on
•
Router
may
have
to
forward
more
than
one
copy
of
packet
•
Can
be
selec;vely
applied
to
logical
connec;ons
so
that
the
flow
from
one
node
to
the
next
–
Does
not
react
to
network
failure
or
conges;on
•
The
OSPF
protocol
is
an
example
•
Need
conven;on
to
iden;fy
mul;cast
addresses
(IPv4,
Class
D,
IPv6)
is
only
restricted
or
halted
on
some
connec;ons
Fast
Retransmit
•
The
second-‐genera;on
rou;ng
algorithm
for
ARPANET
also
uses
this
approach
–
In
IPv4,
1110
as
their
high-‐order
4
bits,
followed
by
a
28-‐bit
group
iden;fier.
•
Under
some
circumstances
improve
on
the
performance
provided
by
RTO
Rou;ng
Strategies
-‐
Flooding
–
In
IPv6,
a
128-‐bit
mul;cast
address
consists
of
an
8-‐bit
prefix
of
all
ones,
a
4-‐bit
flags
field,
a
Choke
Packet
•
Takes
advantage
of
the
rule
that
if
a
TCP
en;ty
receives
a
segment
out
of
order,
it
must
•
Packet
sent
by
node
to
every
neighbor
•
Eventually
mul;ple
copies
arrive
at
des;na;on
•
Path-‐Vector
Rou;ng
4-‐bit
scope
field,
and
a
112-‐bit
group
iden;fier
•
A
control
packet
immediately
issue
an
ACK
for
the
last
in-‐order
segment
that
was
received
No
network
informa;on
required
•
Each
packet
is
uniquely
numbered
so
duplicates
can
be
•
Alterna;ve
to
dispense
with
rou;ng
metrics
and
simply
provide
•
Nodes
translate
between
IP
mul;cast
addresses
and
list
of
networks
containing
group
–
Generated
at
congested
node
–
Sent
back
to
source
node
•
TCP
sender
wait
un;l
it
receives
three
duplicate
ACKs
to
the
same
segment
discarded
•
Need
to
limit
incessant
retransmission
of
packets
informa;on
about
which
networks
can
be
reached
by
a
given
router
and
the
ASs
visited
in
order
to
reach
members
Ø
An
example
is
the
Internet
Control
Message
Protocol
(ICMP)
Source
•
It
is
highly
likely
that
the
following
segment
has
been
lost
and
should
be
retransmiJed
–
Nodes
can
remember
iden;ty
of
packets
retransmiJed
–
Can
include
a
hop
count
in
the
des;na;on
network
by
this
route
–
allows
the
node
to
construct
a
shortest-‐path
spanning
tree
to
all
of
the
networks
Quench
packet
immediately,
rather
than
wai;ng
for
a
;me-‐out.
packets
•
Differs
from
a
distance-‐vector
algorithm
in
two
respects:
containing
group
members.
•
From
router
or
des;na;on
end
system
•
Source
cuts
back
un;l
it
no
longer
receives
quench
–
The
path-‐vector
approach
does
not
include
a
distance
or
cost
es;mate
•
Router
must
translate
between
IP
mul;cast
address
and
network
mul;cast
address
messages
•
Message
is
issued
for
every
discarded
packet
•
Message
may
also
be
issued
for
Fast
Recover
Proper;es
of
Flooding
–
Each
block
of
rou;ng
informa;on
lists
all
of
the
ASs
visited
in
order
to
reach
the
des;na;on
network
–
A
router
aJached
to
an
IEEE
802
network
must
translate
a
32-‐bit
IPv4
or
a
128-‐bit
IPv6
an;cipated
conges;on
•
Retransmit
the
lost
segment,
cut
cwnd
in
half,
and
then
proceed
with
the
linear
increase
All
possible
routes
are
tried
by
this
route
mul;cast
address
into
a
48-‐bit
IEEE
802
MAC-‐level
mul;cast
address.
•
Is
a
crude
technique
for
controlling
conges;on
of
cwnd
Highly
robust
•
The
path
informa;on
enables
a
router
to
perform
policy
rou;ng
•
RFC
3782
modifies
the
fast
recovery
algorithm
to
improve
the
response
when
two
segments
Can
be
used
to
send
emergency
messages
–
A
router
may
decide
to
avoid
a
par;cular
path
in
order
to
avoid
transi;ng
a
par;cular
AS
Requirements
for
Mul;cas;ng
Implicit
Conges7on
Signaling
are
lost
within
a
single
window
At
least
one
packet
will
have
taken
minimum
hop
route
•
For
quality,
performance,
or
security
reasons.
•
Mechanism
required
for
hosts
to
join
and
leave
mul;cast
group
•
Routers
must
exchange
•With
network
conges;on:
Nodes
directly
or
indirectly
connected
to
source
are
visited
–
Minimize
transit
ASs
informa;on
-‐Transmission
delay
increases
l
Packets
may
be
discarded
Explicit
Conges7on
No7fica7on
(ECN)
Disadvantages:
–
Which
networks
include
members
of
given
group
–
Sufficient
informa;on
to
work
out
•Source
can
detect
conges;on
and
reduce
flow
ØResponsibility
of
end
systems
ØEffec;ve
on
An
example
of
explicit
conges;on
signaling
is
the
ECN
capability
High
traffic
load
generated
Border
Gateway
Protocol
(BGP)
shortest
path
to
each
network
connec;onless
(datagram)
networks
•Also
used
in
connec;on-‐oriented
networks
provided
in
IP
and
TCP
and
defined
in
RFC
3168
Security
concerns
•
Was
developed
for
use
in
conjunc;on
with
internets
that
•
Rou;ng
algorithm
to
calculate
shortest
path
-‐LAPF
control
is
capable
of
detec;ng
lost
frames
employ
the
TCP/IP
suite
•
Routers
must
determine
rou;ng
paths
based
on
source
and
des;na;on
addresses
IP
header
includes
the
2-‐bit
ECN
field,
which
enables
routers
to
indicate
to
end
nodes
packets
Rou;ng
Strategies
-‐
Random
Rou;ng
•
Has
become
the
preferred
exterior
router
protocol
for
the
Internet
•Backward
that
are
experiencing
conges;on,
without
the
necessity
of
immediately
dropping
such
packets
•
Simplicity
of
flooding
with
much
less
traffic
load
•
Designed
to
allow
routers
in
different
autonomous
systems
to
cooperate
in
the
exchange
of
rou;ng
Internet
Group
Management
Protocol
(IGMP)
-‐Conges;on
avoidance
no;fica;on
in
opposite
direc;on
to
packet
required
•
A
value
of
00
indicates
a
packet
that
is
not
using
ECN
•
Node
selects
one
outgoing
path
for
retransmission
of
incoming
packet
informa;on
•
Defined
in
RFC
3376
-‐Backward
informa;on
is
transmiJed
either
by
altering
bits
in
a
header
of
a
data
packet
•
As
value
of
01
or
10
is
set
by
the
data
sender
to
indicate
that
the
end-‐points
of
the
transport
•
Selec;on
can
be
random
or
round
robin
•
Protocol
operates
in
terms
of
messages,
which
are
sent
over
TCP
connec;ons
•
Used
to
exchange
mul;cast
group
informa;on
between
hosts
and
routers
on
a
LAN
headed
for
the
source
to
be
controlled
or
by
transmiqng
separate
control
packets
to
the
protocol
are
ECN-‐capable
•
A
refinement
is
to
select
outgoing
path
based
on
probability
calcula;on
•
Current
version
is
known
as
BGP-‐4
(RFC
4271)
•
Hosts
send
messages
to
routers
to
subscribe
and
unsubscribe
from
mul;cast
group
source.
•
A
value
of
11
is
set
by
a
router
to
indicate
conges;on
has
been
encountered
•
No
network
informa;on
needed
Three
func;onal
procedures:
•
Routers
check
which
mul;cast
groups
are
of
interest
to
which
hosts
•Forward
•
Random
route
is
typically
neither
least
cost
nor
minimum
hop
Neighbor
acquisi;on
Neighbor
reachability
Network
reachability
•
IGMP
currently
at
version
3
-‐Conges;on
avoidance
no;fica;on
in
same
direc;on
as
packet
required
TCP
header
includes
two
one-‐bit
flags:
-‐This
informa;on
may
be
transmiJed
either
as
altered
bits
in
data
packets
or
in
separate
•
ECN-‐Echo
flag
•
CWR
(conges;on
window
reduced)
flag
Rou;ng
Strategies
-‐
Adap;ve
Rou;ng
BGP-‐4
Messages
Opera;on
of
IGMP
v1
and
v2
control
packets
•
Used
by
almost
all
packet
switching
networks
•
Rou;ng
decisions
change
as
condi;ons
on
Open
•
IGMPv1
-‐End
system
echoes
the
signal
back
along
the
logical
connec;on
to
the
source
TCP
Friendliness
the
network
Used
to
open
a
neighbor
rela;onship
with
another
router.
–
Hosts
could
join
group
•
Also
known
as
TCP
compa;ble,
refers
to
flows
that
will
back
off
appropriately
in
the
face
of
change
due
to
failure
or
conges;on
•
Requires
informa;on
about
network
Update
–
Routers
used
;mer
to
unsubscribe
members
Explicit
Signaling
Categories
conges;on
–
informa;on
about
the
state
of
the
network
must
be
exchanged
among
the
nodes
Used
to
(1)
transmit
informa;on
about
a
single
route
and/or
(2)
list
mul;ple
routes
to
be
withdrawn.
•
IGMPv2
enabled
hosts
to
unsubscribe
•Binary
•
Conges;on
collapse
Disadvantages:
Decisions
more
complex
Keepalive
•
Opera;onal
model:
–
Receivers
have
to
subscribe
to
groups
–
Sources
do
not
have
to
-‐A
bit
set
in
a
packet
indicates
conges;on
–
Too
many
traffic
flows
opera;ng
independently
and
without
regard
to
their
effect
on
overall
Tradeoff
between
quality
of
network
informa;on
and
overhead
Used
to
(1)
acknowledge
an
Open
message
and
(2)
periodically
confirm
the
neighbor
rela;onship.
subscribe
to
•Credit
based
Internet
performance
Reac;ng
too
quickly
can
cause
oscilla;on
Reac;ng
too
slowly
means
informa;on
may
be
No;fica;on
Send
when
an
error
condi;on
is
detected.
groups
-‐Indicates
how
many
packets
source
may
send
Common
for
end-‐to-‐end
flow
control
–
With
the
ever-‐increasing
load
from
mul;media
applica;ons,
much
of
the
traffic
on
the
irrelevant
–
Any
host
can
send
traffic
to
any
mul;cast
group
•Rate
based
Internet
use
UDP
or
some
other
connec;onless
protocol.
•
Problems:
-‐Supply
explicit
data
rate
limit
Nodes
along
path
may
request
rate
reduc;on
–
These
applica;ons
generally
do
not
take
into
considera;on
overall
Internet
conges;on
Adap;ve
Rou;ng
Advantages
Neighbor
Acquisi;on
–
Spamming
of
mul;cast
groups
effects
and
hence
can
contribute
to
conges;on
collapse.
Improved
performance
•
Occurs
when
two
neighboring
routers
in
different
autonomous
systems
agree
to
exchange
rou;ng
–
Establishment
of
distribu;on
trees
is
problema;c
Aid
in
conges;on
control
informa;on
regularly
•
This
is
mainly
because
the
loca;on
of
sources
is
not
known
Traffic
Management
DCCP
Opera7on
These
benefits
depend
on
the
soundness
of
the
design
and
nature
of
the
load
•
Two
routers
send
Open
messages
to
each
other
aher
a
TCP
connec;on
is
established
–
Finding
globally
unique
mul;cast
addresses
is
difficult
•
Fairness
•Datagram
Conges;on
Control
Protocol
–
If
each
router
accepts
the
request,
it
returns
a
Keepalive
message
in
response
•
It
is
always
possible
that
another
mul;cast
group
-‐Provide
equal
treatment
of
various
flows
•Defined
in
RFC
4340
Classifica;on
of
Adap;ve
Rou;ng
Strategies
•
Protocol
does
not
address
the
issue
of
how
one
router
knows
the
address
or
even
the
existence
of
uses
the
same
mul;cast
address.
•
Quality
of
service
•Connec;on-‐oriented
•
A
convenient
way
to
classify
is
on
the
basis
of
informa;on
source
another
router
nor
how
it
decides
that
it
needs
to
exchange
rou;ng
informa;on
with
that
par;cular
-‐Different
treatment
for
different
connec;ons
•Runs
on
top
of
IP
and
serves
as
an
alterna;ve
transport
protocol
for
applica;ons
that
would
•
Route
to
outgoing
link
with
shortest
queue
•
Can
include
bias
for
each
des;na;on
•
Rarely
router
IGMP
v3
•
Reserva;ons
otherwise
use
UDP
used
-‐
does
not
make
use
of
available
informa;on
•
Addresses
weaknesses
by:
-‐Traffic
contract
between
user
and
network
Local
(isolated)
Open
Shortest
Path
First
(OSPF)
Protocol
–
Allowing
hosts
to
specify
list
from
which
they
want
to
receive
traffic
–
Blocking
traffic
from
-‐Excess
traffic
discarded
or
handled
on
a
best-‐effort
basis
DCCP
Packet
Types
•
Takes
advantage
of
delay
and
outage
informa;on
•
Distributed
or
centralized
•
RFC
2328
•
Used
as
the
interior
router
protocol
in
TCP/IP
networks
•
Computes
a
route
through
the
other
hosts
at
routers
–
Allowing
hosts
to
block
packets
from
sources
that
send
unwanted
DCCP-‐Request:
Adjacent
nodes
internet
that
incurs
the
least
cost
based
on
a
user-‐
traffic
Traffic
Shaping/Traffic
Policing
Sent
by
the
client
to
ini;ate
a
connec;on
(the
first
part
of
the
three-‐way
ini;a;on
handshake)
•
Like
adjacent
configurable
metric
of
cost
•
Is
able
to
equalize
loads
over
mul;ple
equal-‐cost
paths
•
Two
important
tools
in
network
management:
DCCP-‐Response:
All
nodes
•
Each
router
maintains
a
database
that
reflects
the
known
topology
of
the
AS
of
which
it
is
a
part.
IGMP
Opera;on
-‐
Joining
–
Traffic
shaping
Sent
by
the
server
in
response
to
a
DCCP-‐Request
(the
second
part
of
the
three-‐way
ini;a;on
•
The
topology
is
expressed
as
a
directed
graph.
•
IGMP
host
wants
to
make
itself
known
as
group
member
to
other
hosts
and
routers
on
LAN
•-‐Concerned
with
traffic
leaving
the
switch
•
Reduces
packet
clumping
•
Produces
an
output
handshake)
ARPANET
Rou;ng
Strategies
1st
Genera;on
•
IGMPv3
can
signal
group
membership
with
filtering
capabili;es
with
respect
to
sources
packet
stream
that
is
less
burst
and
with
a
more
DCCP-‐Data:
•
Distance
vector
rou;ng
-‐
1969
•
Distributed
adap;ve
using
es;mated
delay
The
graph
consists
of
the
following:
•
Ver;ces,
or
nodes,
of
two
types:
–
EXCLUDE
mode
–
all
members
except
those
listed
–
INCLUDE
mode
–
only
from
group
regular
flow
of
packets
Used
to
transmit
applica;on
data
–
Queue
length
used
as
es;mate
of
delay
1.
Router
2.
Network,
which
is
in
turn
of
two
types
1.
transit,
if
it
can
carry
data
that
neither
originate
members
listed
•Traffic
policing
DCCP-‐Ack:
•
Version
of
Bellman-‐Ford
algorithm
•
Node
exchanges
delay
vector
with
neighbors
•
nor
terminate
on
an
end
system
aJached
to
this
To
join
a
group
a
host
sends
an
IGMP
membership
report
message
-‐Concerned
with
traffic
entering
the
switch
Used
to
transmit
pure
acknowledgments
Update
rou;ng
table
based
on
incoming
informa;on
•
Doesn't
consider
line
speed,
just
network
•
Address
field
is
the
mul;cast
address
of
group
•
Sent
in
an
IP
datagram
with
the
same
-‐Packets
that
don’t
conform
may
be
treated
in
one
of
the
following
ways:
DCCP-‐DataAck:
queue
length
and
2.
stub,
if
it
is
not
a
transit
network
mul;cast
des;na;on
address
•
Current
group
members
receive
and
learn
new
member
•
•-‐Give
the
packet
lower
priority
compared
to
packets
in
other
output
queues
•
Label
the
Used
to
transmit
applica;on
data
with
piggybacked
acknowledgment
informa;on
responds
slowly
to
conges;on
•
Edges
are
of
two
types:
Routers
listen
to
all
IP
mul;cast
addresses
to
hear
all
reports
packet
as
nonconforming
by
seqng
the
appropriate
bits
in
a
header•
Discard
the
packet
DCCP-‐CloseReq:
1.
Graph
edges
that
connect
two
router
ver;ces
when
the
corresponding
routers
are
connected
to
each
Sent
by
the
server
to
request
that
the
client
close
the
connec;on
Since
queue
lengths
vary
rapidly
with
;me,
the
distributed
percep;on
of
the
shortest
route
other
by
a
direct
IGMP
Opera;on
Keeping
Lists
Valid
Token
Bucket
DCCP-‐Close:
could
change
while
a
packet
is
en
route.
This
could
lead
to
a
thrashing
situa;on
in
which
a
point-‐to-‐point
link
Routers
periodically
issue
IGMP
general
query
message
•
Widely
used
traffic
management
tool
•
Advantages:
Used
by
the
client
or
the
server
to
close
the
connec;on;
elicits
a
DCCP-‐Reset
in
response
packet
con;nues
to
seek
out
areas
of
low
conges;on
rather
than
aiming
at
the
des;na;on.
2.
Graph
edges
that
connect
a
router
vertex
to
a
network
vertex
when
the
router
is
directly
connected
•
In
datagram
with
all-‐hosts
mul;cast
address
–
Many
traffic
sources
can
be
defined
easily
and
accurately
DCCP-‐Reset:
to
the
network
•
Hosts
must
read
such
datagrams
–
Provides
a
concise
descrip;on
of
the
load
to
be
imposed
by
a
flow,
enabling
the
service
to
Used
to
terminate
the
connec;on,
either
normally
or
abnormally
ARPANET
Rou;ng
Strategies
2nd
Genera;on
•
Hosts
respond
with
report
message
determine
easily
the
resource
requirement
DCCP-‐Sync,
DCCP-‐SyncAck:
•
Link-‐State
Rou;ng
-‐
1979
•
Distributed
adap;ve
using
delay
criterion
Two
routers
joined
by
a
point-‐to-‐point
link
are
represented
in
the
graph
as
a
pair
of
edges
(e.g.,
routers
Router
doesn’t
know
every
host
in
a
group
–
Provides
the
input
parameters
to
a
policing
func;on
Used
to
resynchronize
sequence
numbers
aher
large
bursts
of
loss
–
Using
;mestamps
of
arrival,
departure
and
ACK
;mes
6
and
10).
•
When
mul;ple
routers
are
aJached
to
a
network
(such
as
a
LAN
or
packet-‐switching
•
Needs
to
know
at
least
one
group
member
s;ll
ac;ve
•
Re-‐computes
average
delays
every
10
seconds
•
Any
changes
are
flooded
to
all
other
network),
the
directed
graph
shows
all
routers
bidirec;onally
connected
to
the
network
vertex
(e.g.,
•
Each
host
in
group
sets
;mer
with
random
delay
During
any
;me
period
T,
the
amount
of
data
sent
cannot
exceed
RT
+
B.
DCCP
Conges;on
Control
nodes
•
Re-‐computes
rou;ng
using
Dijkstra’s
algorithm
•
Good
under
light
and
medium
routers
1,
2,
3,
and
4
all
connect
to
network
3).
•
If
a
single
router
is
aJached
to
a
network,
the
network
•
Host
hearing
another
report
cancels
own
•
If
;mer
expires,
host
sends
report
•
Only
one
Token
bucket
is
typically
used
for
traffic
shaping
but
not
traffic
policing.
•
The
Internet
DCCP
Working
Group
has
defined
two
conges;on
control
mechanisms,
each
loads
•
Under
heavy
loads,
liJle
correla;on
between
reported
delays
will
appear
in
the
graph
as
a
stub
connec;on
(e.g.,
network
7).
•
A
host
can
be
directly
connected
to
a
member
of
each
group
reports
to
assigned
a
conges;on
control
iden;fier
(CCID):
and
those
experienced
router,
in
which
case
it
is
depicted
in
the
corresponding
graph
(e.g.,
host
1).
•
If
a
router
is
connected
to
router
Comparison
TCP-‐like
conges;on
control
(CCID
2)
other
autonomous
systems,
then
the
path
cost
to
each
network
in
the
other
system
must
be
obtained
•
A
token
bucket
fills
at
a
constant
rate
up
to
the
capacity
of
the
bucket
and
emp;es
at
a
rate
•
This
mechanism
has
behavior
modeled
directly
on
TCP,
including
conges;on
window,
slow
Packet-‐Switching
Network
Subject
to
Oscilla;ons
by
some
exterior
router
protocol.
Each
such
network
is
represented
on
the
graph
by
a
stub
and
an
edge
IGMP
Opera;on
-‐
Leaving
dictated
by
the
input
data
stream
while
the
bucket
is
not
empty
start,
and
;me-‐outs
•
A
significant
por;on
of
available
capacity
is
unused
at
just
the
;me
when
it
is
needed
most
to
the
router
with
the
known
path
cost
(e.g.,
networks
12
through
15).
•
Host
leaves
group
by
sending
a
leave
group
message
to
the
all-‐routers
sta;c
mul;cast
•
A
leaky
bucket
emp;es
at
a
constant
rate
while
the
bucket
is
not
empty
and
fills
at
a
rate
•
Applica;ons
should
use
CCID
2
if
they
prefer
maximum
bandwidth
u;liza;on
to
steadiness
•
The
overu;liza;on
of
some
links
can
lead
to
the
spread
of
conges;on
within
the
network
•
address
dictated
by
the
input
stream
up
to
the
capacity
of
the
bucket.
of
rate
The
large
swings
in
measured
delay
values
result
in
the
need
for
more
frequent
rou;ng
Dijkstra’s
Algorithm
–
Sends
a
membership
report
message
with
EXCLUDE
op;on
and
null
list
of
source
addresses
•
For
Token
bucket,
as
the
rate
of
incoming
packets
rise,
the
output
of
the
system
speeds
up.
TCP-‐friendly
rate
control
(TFDC)
(CCID
3)
update
messages.
•
This
increases
the
load
on
the
network
at
just
the
;me
when
the
•
Finds
shortest
paths
from
given
source
nodes
to
all
other
nodes
•
Develop
paths
in
order
of
increasing
•
Router
determines
if
any
group
members
using
group-‐specific
query
message
remain
–
In
effect,
token
bucket
gives
credit
to
a
flow
or
connec;on
that
is
underused,
up
to
a
point.
•
This
mechanism
is
based
on
the
use
of
a
variant
of
the
TCP-‐friendly
equa;on
network
is
already
stressed.
path
length
•
Algorithm
runs
in
stages
•
Suitable
for
applica;ons
such
as
streaming
media,
where
a
rela;vely
smooth
sending
rate
is
–
Each
;me
adding
node
with
next
shortest
path
Group
Membership
with
IPv6
Conges7on
Control
in
Packet-‐Switching
Networks
important
ARPANET
Rou;ng
Strategies
3rd
Genera;on
•
Algorithm
terminates
when
all
nodes
have
been
added
to
T
•
IGMP
defined
for
IPv4
Send
control
packet
to
some
or
all
source
nodes
•
1987
–
Uses
32-‐bit
addresses
•
Requires
addi;onal
traffic
during
conges;on
•
Link
cost
calcula;on
changed
Dijkstra’s
Algorithm
Method
•
IPv6
internets
need
func;onality
Rely
on
rou;ng
informa;on
–
Damp
rou;ng
oscilla;ons
–
Reduce
rou;ng
overhead
Step
3
[Update
Least-‐Cost
Paths]
•
IGMP
func;ons
included
in
Internet
Control
Message
Protocol
v6
(ICMPv6)
•
link
delay
informa;on
Rou;ng
in
Packet
Switching
Networks
•
Measure
average
delay
over
last
10
seconds
and
transform
into
link
u;liza;on
es;mate
L(n)
=
min[L(n),
L(x)
+
w(x,
n)]
for
all
n
Ï
T
–
ICMPv6
has
func;onality
of
ICMPv4
&
IGMP
•
May
react
too
quickly
•
Key
design
issue
for
(packet)
switched
networks
•
Select
route
across
network
•
Normalize
this
based
on
current
value
and
previous
results
If
laJer
term
is
minimum,
path
from
s
to
n
is
path
from
s
to
x
concatenated
with
edge
from
x
to
n
•
ICMPv6
includes
group-‐membership
query
and
group-‐membership
report
message
End
to
end
probe
packets
between
end
nodes
•
Characteris;cs
required:
•
Set
link
cost
as
func;on
of
average
u;liza;on
Step
2
[Get
Next
Node]
•
packet
;me
stamped
to
measure
the
delay
between
two
endpoints
–
Correctness
–
Simplicity
–
Robustness
–
Stability
–
Fairness
–
Op;mality
–
Efficiency
Find
neighboring
node
not
in
T
with
least-‐cost
path
from
s
Protocol
Independent
Mul;cast
(PIM)
•
Adds
to
overhead
Internet
Rou;ng
Protocols
Incorporate
node
into
T
•
A
separate
rou;ng
protocol,
independent
of
any
exis;ng
unicast
rou;ng
protocol
Add
conges;on
informa;on
topackets
in
transit
CSE4344
–
Computer
Networks
and
Distributed
Systems
•
Routers
are
responsible
for
receiving
and
forwarding
packets
through
the
interconnected
Also
incorporate
the
edge
that
is
incident
on
that
node
and
a
node
in
T
that
contributes
to
the
path
•
Designed
to
extract
needed
rou;ng
informa;on
from
any
unicast
rou;ng
protocol
•
Either
backwards
or
forwards
Rou;ng
set
of
networks
Step
1
[Ini;aliza;on]
•
Recognizes
that
a
different
approach
may
be
needed
to
mul;cast
rou;ng
depending
on
the
Dr.
Padmaraj
Nair
–
Makes
rou;ng
decisions
based
on
knowledge
of
the
topology
and
traffic/delay
T
=
{s}
Set
of
nodes
so
far
incorporated
concentra;on
of
mul;cast
group
members
Retransmission
Timer
Management
condi;ons
of
the
internet
–
Routers
exchange
rou;ng
informa;on
using
a
special
rou;ng
L(n)
=
w(s,
n)
for
n
≠
s
As
network
or
internet
condi;ons
change
a
Rou;ng
in
Packet
Switching
Networks
protocol
Ini;al
path
costs
to
neighboring
nodes
are
simply
link
costs
PIM
sta;c
retransmission
;mer
is
likely
to
be
either
too
long
or
too
short
•
Key
design
issue
for
(packet)
switched
networks
•
Select
route
across
network
Ø
Two
concepts
in
considering
the
rou;ng
func;on:
Dense-‐Mode
PIM
Virtually
all
TCP
implementa;ons
aJempt
to
es;mate
the
current
round-‐trip
;me
and
then
between
end
nodes
•
Characteris;cs
required:
–
Rou;ng
informa;on
Bellman-‐Ford
Algorithm
•
Appropriate
for
intra-‐AS
mul;cast
rou;ng
set
the
;mer
to
a
value
somewhat
greater
than
the
es;mated
;me
–
Correctness
–
Simplicity
–
Robustness
–
Stability
–
Fairness
–
Op;mality
–
Efficiency
Ø
Informa;on
about
the
topology
and
delays
of
the
internet
•
Find
shortest
paths
from
given
node
subject
to
constraint
that
paths
contain
at
most
one
link
•
May
be
viewed
as
a
poten;al
alterna;ve
to
a
mul;cast
version
of
OSPF
known
as
MOSPF
Simple
average:
–
Rou;ng
algorithm
•
Find
the
shortest
paths
with
a
constraint
of
paths
of
at
most
two
links
•
Proceeds
in
stages
•
When
there
are
many
mul;cast
members
within
a
configura;on,
then
the
frequent
•
Take
the
average
of
observed
round-‐trip
;mes
over
a
number
of
segments
Elements
of
Rou;ng
Techniques
for
Packet-‐Switching
Networks
Ø
The
algorithm
used
to
make
a
rou;ng
decision
for
a
par;cular
datagram,
based
on
exchange
of
group
membership
informa;on
•
If
the
average
accurately
predicts
future
round-‐trip
;mes,
then
the
resul;ng
retransmission
Performance
Criteria
current
rou;ng
informa;on
Bellman-‐Ford
Algorithm
is
jus;fied
;mer
will
yield
good
performance
Number
of
hops
Cost
Delay
Step1
[Ini;aliza;on]
L0(n)
=
¥,
for
all
n
¹
s
Lh(s)
=
0,
for
all
h
•
It
is
desirable
to
build
shared
spanning
trees
so
that
packet
duplica;on
occurs
as
Exponen;al
average:
Throughput
Autonomous
Systems
(AS)
Step
2
[Update]
For
each
successive
h
³
0
For
each
n
≠
s,
compute:
infrequently
as
possible.
•
Technique
for
predic;ng
the
next
value
on
the
basis
of
a
;me
series
of
past
values
Decision
Time
•
Exhibits
the
following
characteris;cs:
Lh+1(n)=min
•
Specified
in
RFC
793
Packet
(datagram)
Session
(virtual
circuit)
–
Is
a
set
of
routers
and
networks
managed
by
a
single
organiza;on
–
Consists
of
a
group
of
j[Lh(j)+w(j,n)]
Sparse-‐Mode
PIM
Decision
Place
routers
exchanging
informa;on
via
a
common
rou;ng
protocol
–
Except
in
;mes
of
failure,
is
Connect
n
with
predecessor
node
j
that
gives
min
•
Suited
for
inter-‐AS
mul;cast
rou;ng
•
Defined
as
a
group
in
which:
Exponen7al
RTO
Backoff
Each
node
(distributed)
Central
node
(centralized)
Origina;ng
node
(source)
connected
(in
a
graph-‐theore;c
sense);
there
is
a
path
between
any
Eliminate
any
connec;on
of
n
with
different
predecessor
node
formed
during
an
earlier
itera;on
–
The
number
of
networks/domains
with
group
members
present
is
significantly
smaller
than
•
When
a
TCP
sender
;mes
out
on
a
segment,
it
must
retransmit
that
segment
Network
Informa;on
Source
pair
of
nodes
Path
from
s
to
n
terminates
with
link
from
j
to
n
the
number
of
networks/domains
in
the
internet
–
RFC
793
assumes
that
the
same
RTO
value
will
be
used
None
Local
Adjacent
node
–
The
internet
spanned
by
the
group
is
not
sufficiently
resource
rich
to
ignore
the
overhead
of
–
Because
the
;me-‐out
is
probably
due
to
network
conges;on,
maintaining
the
same
RTO
Nodes
along
route
All
nodes
Interior
Router
Protocol
(IRP)
current
mul;cast
rou;ng
schemes
value
is
ill
advised
Network
Informa;on
Update
Timing
•
A
shared
rou;ng
protocol
which
passes
rou;ng
informa;on
between
routers
within
an
AS
Comparison
•
Flooding
of
mul;cast
group
info
to
all
routers
is
inefficient,
because
most
routers
will
not
be
•
A
more
sensible
policy
dictates
that
a
sending
TCP
en;ty
increase
its
RTO
each
;me
a
Con;nuous
Periodic
Major
load
change
•
Custom
tailored
to
specific
applica;ons
and
requirements
•
Bellman-‐Ford
along
the
path
of
any
members
of
a
given
mul;cast
group
segment
is
retransmiJed
Topology
change
–
Calcula;on
for
node
n
needs
link
cost
to
neighboring
nodes
plus
total
cost
to
each
neighbor
from
s
•
LiJle
opportunity
for
using
shared
spanning
trees,
and
the
focus
should
be
on
providing
Exterior
Router
Protocol
(ERP)
–
Each
node
can
maintain
set
of
costs
and
paths
for
every
other
node
mul;ple
shortest-‐path
unicast
routes
Karn’s
Algorithm
•
Protocol
used
to
pass
rou;ng
informa;on
between
routers
in
different
ASs
–
Can
exchange
informa;on
with
direct
neighbors
•Do
not
use
the
measured
RTT
for
a
retransmiJed
segment
to
update
SRTT
and
SDEV
Performance
Criteria
•
Will
need
to
pass
less
informa;on
than
an
IRP
for
the
following
reason:
–
Can
update
costs
and
paths
based
on
informa;on
from
neighbors
and
knowledge
of
link
costs
PIM
•Calculate
the
backoff
RTO
using
the
equa;on
RTO
=
q
*
RTO
when
a
retransmission
occurs
•
Used
for
selec;on
of
route
•
Simplest
is
to
choose
“minimum
hop”
•
Can
be
–
If
a
datagram
is
to
be
transferred
from
a
host
in
one
AS
to
a
host
in
another
AS,
a
router
in
•
Dijkstra
•
For
a
mul;cast
group,
one
router
is
designated
as
a
rendezvous
point
(RP).
•Use
the
backoff
RTO
value
for
succeeding
segments
un;l
an
acknowledgment
arrives
for
a
generalized
as
“least
cost”
rou;ng
•
Because
“least
cost”
is
more
flexible
it
is
more
the
first
system
need
only
determine
the
target
AS
and
devise
a
route
to
get
into
that
target
–
Each
node
needs
complete
topology
–
The
selec;on
of
an
RP
for
a
given
mul;cast
group
is
a
dynamic
process
segment
that
has
not
been
retransmiJed
common
than
“minimum
hop”
system
–
Must
know
link
costs
of
all
links
in
network
•
A
group
des;na;on
router
sends
a
Join
message
toward
the
RP
–
Once
the
datagram
enters
the
target
AS,
the
routers
within
that
–
Must
exchange
informa;on
with
all
other
nodes
–
The
reques;ng
router
uses
a
unicast
shortest-‐path
route
to
transmit
the
message
toward
system
can
cooperate
to
deliver
the
datagram
the
RP.
–
The
reverse
of
this
path
becomes
part
of
the
distribu;on
tree
from
this
RP
to
–
The
ERP
is
not
concerned
with,
and
does
not
know
about,
the
details
of
the
route
Evalua;on
listeners
in
this
group.
Dependent
on
•
Any
node
that
wishes
to
send
to
a
mul;cast
group
sends
packets
toward
the
RP,
using
a
•
Processing
;me
of
algorithms
shortest-‐path
unicast
route.
•
Amount
of
informa;on
required
from
other
nodes
•
From
the
RP,
transmission
occurs
down
the
tree
to
the
listeners,
with
each
packet
Implementa;on
specific
replicated
at
each
split
in
the
tree.
Both
converge
under
sta;c
topology
and
costs
•
In
a
widely
dispersed
group,
any
RP
will,
of
necessity,
be
remote
from
many
of
the
group
Both
converge
to
same
solu;on
members,
and
paths
for
many
group
members
will
be
much
longer
than
the
least-‐cost
path.