FIT1047 Week 9
Networks: Network and Transport layers
FIT1047
Goals for this week
• See how routers connect different networks
• Understand how the transport layer makes sure messages arrive correctly and at the right process
• Study the structure of the Internet
FIT1047
2
The Network Layer:
Addresses
FIT1047
The Network Layer: Addresses
What is Addressing in a layered architecture?
Any communication that requires two parties (any two from a larger pool of entities), needs a mechanism to specify the source address and destination address. As we have logical connections, we have various forms of addressing in each layer.
FIT1047
Addressing devices
192.168.1.15
192.168.1.1
216.58.220.142 216.239.46.151
209.85.142.94
203.214.44.94
202.158.200.149
202.158.200.12 202.158.210.41
Each “layer 3 device” has an address!
Routers decide path for packets that need to leave their own network.
FIT1047
5
IP version 4 addresses
32 bit addresses
• Written using “dotted decimal” notation • Example: 130.194.66.43
10000010 11000010
01000010
LAN
00101011 Host
Monash University
S
)
N
)S )
N)
N
e
e
t
tw
w
o
Hierarchy used for routing
o
r
r
k
k(
(
M
Mo
on
na
a
s
sh
h
• You can immediately see if a destination address is in
u
ub
bn
ne
e
t
t(
(
L
L
A
AN
the same subnet!
• Subnet mask: 255.255.255.192 or /26
.o
n
no
o
.
of
f
b
bi
i
t
t
s
s
si
su
u
b
i
n
nn
bn
ne
ne
et
et
t
tw
wo
or
rk
k+
+
FIT1047
6
IP version 4 addresses
Example: IP address PRIVATE Class C = 192.168.1.0 with regular mask =255.255.255.0
With two bit of subnet the classless Subnet mask = 255.255.255.192 ( 11111111. 11111111.11111111.11000000 ) Subnet Bits = 2
Maximum Subnets = 4
# Network ID
0 192.168.1.0
1 192.168.1.64
2 192.168.1.128
3 192.168.1.192
Valid IP Range/floor 192.168.1.1 – 192.168.1.62 192.168.1.65 – 192.168.1.126 192.168.1.129 – 192.168.1.190 192.168.1.193 – 192.168.1.254
Broadcast Address
192.168.1.63 192.168.1.127 192.168.1.191 192.168.1.255
FIT1047
Network Classes
Subnet masks often expressed using dotted notation
• e.g. 255.255.0.0 meaning /16
Previously used hierarchy:
• Class A: /8 (e.g. IBM, MIT, AT&T, Apple, …)
• Class B: /16 (e.g. Monash 130.194.0.0/16)
• Class C: /24
Now: classless
• e.g. /22, which can also be written as 255.255.252.0
FIT1047
8
Subnets
130.194.76.192
130.194.76.253 130.194.76.191
Backbone
network 130.194.40.13
G
G
a
G
s
G
a
a
a
t
t
t
e
e
w
te
w
a
ew
wa
ay
ay
yr
yr
ro
ou
ro
u
ou
u
t
t
e
er
rs
t
t
e
e
r
r
s
s
130.194.66.61
130.194.40.22
130.194.66.43
FIT1047
9
IP version 6
IPv4 addresses: 32 bits
• In theory 4,294,467,295 addresses
• In practice probably only half are usable
• We’ve run out of new blocks of IPv4 addresses
IPv6
• “new” version of the IP protocol (from 1998…)
• 128 bits per address
• so that’s four times more, right?
…No! IPv6 addresses is more than 7.9×1028 times as many as IPv4
FIT1047
10
IP version 6
2128=
340,282,366,920,938,463,463,374,607,431,768,211,456
(340 undecillion) IPv6 addresses
FIT1047
11
IP version 6
340,282,366,920,938,463,463,374,607,431,768,211,456
A bit excessive?
• At least 7 addresses for every atom of every person on earth
• 665,570,793,348,866,943,898,599 addresses per square meter of the surface of the earth
Required!
• The huge space is used to create hierarchies
• This makes it easy to assign whole subnets
FIT1047
12
IP version 6 address space
Typical allocation:
IANA + RIR
/23 /32
Company / Organization (16 bits)
/48 /64
Interface ID (64 bits)
/128
ISP (9 bits)
Subnet (16 bits)
232 times more addresses than IPv4 in Interface ID alone! Every company can run more than 65,000 different LANs
FIT1047
13
IP version 6 address space
Typical allocation:
232 times more addresses than IPv4 in Interface ID alone! Every company can run more than 65,000 different LANs
FIT1047
14
IPv6 transition
FIT1047
15
The Network Layer:
Routing
FIT1047
Routers
Routers connect networks
• Internet is a network of networks!
• Most important piece of Internet infrastructure
A router is a layer 3 device
• one IP address per interface, i.e. typically per subnet it is
connected to
• Clients send packets to routers if destination is outside their
own subnet
• Routers use IP address to determine where the packet is sent next
FIT1047
17
Routing tables
For each incoming packet, the router
• looks at the packet’s destination IP address
• consults the routing table:
to which other router should I send a packet for this destination, or can I deliver it directly?
• if destination not in table: send to default gateway
• if no default gateway configured: packet can’t be routed
FIT1047
18
RoutDienst. gNexet xaDemst. Npexlt e
A
A
A
C
d
d
s
e
e AG
st
t
:
:H
H
Dest.
Next
default
C
CD
Dest.
Next
default
F
BEF K
H
G
H
K
F
F
B
G
H
K
D
E
E
G
H
K
G
H
K
D
D
H
E
G
H
C
Dest.
A
Next
C
Dest.
A
Next
d
de
e
s
s
t
t
:
:A
A
FIT1047
19
Types of routing
Static routing
• Network manager prepares fixed routing tables
• Manually updated when the network changes
• Used in simple networks that don’t change a lot
FIT1047
20
Static routing demo
FIT1047
Types of routing
Static routing
• Network manager prepares fixed routing tables
• Manually updated when the network changes
• Used in simple networks that don’t change a lot
Dynamic routing
• Routers exchange information to build routing tables dynamically
• Initial tables can be set up by network managers
FIT1047
23
Q: Switch vs router
What is the difference between a forwarding table (in a switch) and a routing table (in a router)? (multiple correct answers)
1. A forwarding table uses the MAC address, a routing table uses the IP address
2. Only forwarding tables can contain multiple destinations per port
3. Routing tables connect different subnets, forwarding happens in
the same subnet
4. Only forwarding tables change when the network changes
5. Routing tables are set up by hand, forwarding is automatic
FIT1047
24
Dynamic routing algorithms
Distance vector
• Exchange information about distance to destination,
choose shortest route
• EIGRP (Enhanced Interior Gateway Routing Protocol)
• RIP (Routing Information Protocol)
• BGP (Border Gateway Protocol)
Link state
• Exchange information about quality of links, choose fastest route
• OSPF (Open Shortest Path First)
FIT1047
25
Distance vector routing
D
Di
D
Di
r
AN
AG
i
s
st
r
e
ta
ec
ct
or
an
ti
r
k
nc
io
k
G
ce
o
e:
n
n
:
G
:
1
1
h
:i
i
n
n
ho
t
o
p
te
er
p
r
f
f
a
a
c
ce
e1
0 3
23
1 2
N
D
Di
e
e
t
is
tw
st
w
o
ta
10N
K
i
B2E1F
K
or
an
rk
nc
c
kA
e
e:
:
0
0h
h
o
op
p
rf
f
s
s
n
C1 0D
D
Di
ir
re
ec
ct
ti
io
on
:
:
i
i
n
nt
t
e
er
a
a
c
c
e
e
0
0
N
Ne
et
t
w
w
o
N
D
Di
e
et
is
tw
st
wo
t
a
a
or
n
r
k
nc
c
k
K
e
e
:
:2
2h
ho
o
p
p
s
s
D
D
i
ir
re
e
c
c
t
ti
io
on
n:
:
i
in
n
t
te
er
r
f
fa
a
c
c
e
e
0
0
I
I
n
nt
t
e
er
r
f
f
a
ac
ce
en
n
u
um
m
b
be
e
r
r
H
FIT1047
26
Routing Information Protocol (RIP)
Dest. Dist. Interf. Dest. Dist. Interf.
A
B
0
0
0
1
C
G
0
0
0
0
AG
1
BEF
All routers send their routing tables to all other routers.
Tables converge after some time.
H
Dest.
Dist.
Interf.
A
0
0
CK E
0
1
0
2
K
0
3
FIT1047
27
E
D
K
G
H
F
0
0
1
1
1
1
C3
2
3
1
3
3
H
F
A
B
E
K
0
1
D
1
1
2
2
3
0
0
0
0
Routing Information Protocol (RIP)
Distance = hop count
• Max. hop count 15
• Avoids loops
Only used in small networks
• At most 15 hops
• Updates transmit whole routing table
• Can be slow to converge
FIT1047
28
Link-state routing
Routers exchange information about connectivity
• not just routing table (best routes)
• but all of the network it knows about
• use a metric (usually link speed) to describe the quality of each connection
• each router can independently compute best route to every subnet
Most common protocol: OSPF
(video available on Moodle in case you’re interested)
FIT1047
29
Address resolution
FIT1047
Address resolution
Assume we browse to http://www.google.com.au
• We have to translate www.google.com.au into an IP address: 216.58.220.99
• We send a request through the Internet to that IP address
• The router in the LAN of 216.58.220.99 needs to know the MAC address for 216.58.220.99 to deliver the frame
This is known as address resolution.
nslookup command for DNS name resolution, from Domain name to IP address
FIT1047
31
Address resolution: Application Layer
DNS (Domain Name System)
• Application layer protocol for address resolution
• Client sends request to DNS server to get IP address registered for a name
DNS Servers
• Implement a distributed database of names
• Are organised in a hierarchy reflecting the structure of the domain names
FIT1047
32
r
Root Zone
Top-Level Domain (TLD)
ro
oo
a
t
a
ot
DNS
s
s
k
kz
z
.
.a
au
u
.
g.
g.
.
o
o
r
rg
.
o
o
r
rg
.
e
e
d
co
m.
g.
u
du
.
du u.
u.
.
c
o
m
.o
o
r
r
a
g
.e
e
d
.m
m
o
on
na
as
u
e
sk
ke
ed
d
n
ns
s
–
.a
h
-3
a
39
96
6
.
.u
un
ni
im
me
el
l
b
b.
.e
d
d
u
as
u.
s
.a
au
h
u
m
s
h.
Iterative DNS lookup
m
o
on
na
as
h
.e
e
d
du
u
u
u
n
ni
i
m
me
FIT1047
33
i
i
i
m
m
m
m
e
e
e
e
el
l
l
lb
b
b
b
b
.
.
.
.
.
e
e
e
e
d
d
d
du
u
u
u
u
.
.
.
.a
a
a
a
a
u
u
u
u
u
?
?
?
?
?
1
b
12
2
4
4
.
.
4
47
el
7.
l
b
.1
17
70
0
.
.
4
46
6
w
w
w
w
w
w
w
w
w
w
ww
w
w
w
w
.
.
.
.
.u
u
u
u
un
n
n
n
n
i
FIT1047
34
c
o
o
m
.o
n
i
me
e
l
lb
o
b
r
rg
.
e
.
.e
e
d
du
u.
.a
au
1
m
o
r
Root Zone
Top-Level Domain (TLD)
Recursive DNS lookup
r
o
oo
e
on
o
t
d
d
.m
na
t
as
s
h
.e
DNS
.
g.
g.
.
o
or
rg
.o
or
r
g
.e
e
d
du
.
u u.
u.
.c
m.
g.
u
1
2
.a
h
e
un
2
au
n
i
4
4.
i
.
4
u
m
o
on
n
a
a
s
sh
m
h.
d
d
u
u
m
me
4
7
7
el
.
lb
.1
b
w
im
e
ww
ww
w.
.u
un
u? ?I
I
S
SP
P
c
ca
ac
17
ch
70
0
he
.
.4
46
6
Iterative & Recursive DNS Query
• The DNS Query 1 is a Recursive DNS Query, and 8 is the reply for it. Where the local DNS server is involved in all the queries. If the DNS Server doesn’t know the answer to the query to the Client, Then, DNS Server may query other DNS Servers on behalf of the DNS Client.
• The DNS Queries 2, 4 and 6 are Iterative DNS Queries and 3, 5 and 7 are the replies for it respectively. Where the host is involved in all the stages of query.
Reference: omnisecu.com
FIT1047
Address resolution: Data Link Layer
How to find the MAC address for an IP address:
1.2.3.4
?
Arp /? command for ARP MAC address to IP_Address resolution
0
00
0:
:2
2
3
3
:
:a
a
e
e:
1.2.3.1
e
:
e
e7
h
7:
o
o
h
:5
5
h
a
2
2:
as
:
8
8
5
s1
5
1.2.3.5
W
s
b
1.2.3.6
W
W
W
h
h
h
ho
o
o
o
h
h
h
h
a
a
a
a
s
s
s
1
1
1
1
.
.
.
.2
2
2
2
.
.
.
.
3
1.
3
3
3
.
.
2
.
.
.5
2.
5
5
5
.3
?
?
?
3.
.5
5?
?
br
r
o
m
me
Address Resolution Protocol (ARP)
oa
ad
es
dc
ss
c
sa
a
a
as
g
g
s
e
t
t
W
W
h
FIT1047
36
Layers of Abstraction
Switch Router Switch
FIT1047
37
Q: WiFi routers
Your WiFi/ADSL modem at home is often called a WiFi Router. What kind of routing does it do?
1. Static routing, all packets are routed to my ISP.
2. Link state routing, to compensate for the sometimes poor quality of ADSL connections.
3. No routing, the name is just used for marketing reasons.
FIT1047
38
Transport Layer
FIT1047
Transmission Control Protocol (TCP)
Connection-oriented
• A virtual circuit is established between two devices
• To the application it always looks like a point-to-point full duplex connection
• Messages split into segments for transmission Reliable
• Errors are detected and corrected
• Segments are re-assembled in the correct order Used by e.g. HTTP, FTP, SMTP, POP, IMAP, SSH…
FIT1047
40
Addressing applications
W
W
a
a
r
h
hy
rr
a
a
n
r
y
d
i
i
v
ve
nd
d
ei
dn
o
o
e
i
n
no
e
s
n
t
ot
s
t
t
tt
h
h
i
t
h
t
h
he
i
h
e
s
s
b
e
o
e
H
HT
br
ot
r
o
TT
ow
th
TP
w
s
he
er
Pr
r
o
se
r
er
o
n
e
es
sp
rw
ne
wi
e?
po
?
o
in
n
n
n
d
d
s
se
o
o
e
w
w
FIT1047
41
Addressing applications
Ra
130.194.77.3 :57017 7130.194.77.3 :57018
7
130.194.77.3 :57019
t
smtp.monash.edu
t
a
n
n
d
d
o
om
http://www.google.com.au
R
216.58.220.9 :80 9
mc
cl
l
i
ie
en
n
t
tp
po
(
o
r
(2
r
t
25
5
=
t,
,
o
o
n
ne
e
p
pe
er
r
c
c
o
on
nn
n
e
ec
c
t
t
i
io
o
n
n
e
d
ds
=
S
s
S
e
e
M
r
r
v
MT
v
T
P
e
e
r
P
)
rp
)
po
o
r
rt
111.119.8.38:80 http://www.nasa.gov
F
F
i
i
x
xe
ed
(
(8
8
0
ds
se
0=
e
=H
r
rv
v
HT
T
T
e
e
r
r
p
TP
po
P)
)
o
r
rt
7 130.194.11.146:25
F
F
i
ix
x
e
FIT1047
42
One address per layer
Application Layer
• URL (e.g. http://www.csse.monash.edu)
Transport Layer (TCP)
• Port number (e.g. 80 for HTTP)
• identifies the application that handles a message
Network Layer (IP)
• IP address (e.g. 130.194.66.43)
• used for identifying devices across networks
Data Link Layer (Ethernet)
• MAC address (e.g. 00:23:ae:e7:52:85)
• used for sending frames in a LAN
FIT1047
43
TCP ARQ
Error control
• Data Link Layer discards frames that have errors
• Frames may not arrive at all
• But TCP should be a reliable channel!
Solution: Automatic Repeat ReQuest
• Exchange acknowledgements (ACK), letting sender
know that packets were received correctly
• Sender re-transmits if no ACK within certain time
FIT1047
44
TCP ARQ
When we send a TCP packet, it includes two numbers. Sequence number:
• how many bytes we’ve already transmitted (before this one)
Acknowledgement number:
• how many bytes we’ve received from the other side
Sender can therefore check how many bytes have been received correctly!
FIT1047
45
Establishing a connection
Three-way handshake:
• Client sends a SYN packet with random sequence
number A
• Server replies with SYN, ACK, acknowledgement
number A+1, and random sequence number B
• Client sends ACK with sequence number A+1 and
acknowledgement number B+1
• TCP connection is established
Ref: Animation link: https://www.inetdaemon.com/tutorials/internet/tcp/3-way_handshake.shtml
FIT1047
46
Closing a connection
Four-way handshake:
• Computer A (client or server!) sends a FIN packet
• Computer B acknowledges with an ACK
• Computer B sends a FIN packet
• Computer A acknowledges with an ACK
• Can be simplified to three-way (combining a FIN/ACK)
Necessary because TCP is full duplex!
FIT1047
47
3
e
32
3
2
0
32
20
31
05
0
18
5|
6
6|
85
|7
|
7
5
|
7
4
4
3
TCP session
Client
Server
3
|0
0S
2
F
3A
SY
F
AC
I
YN
IN
C
K
N
N
3
(
ld
K
h
K
ha
(f
fu
a
ul
l
3
n
n
d
l
du
–
-w
wa
ds
s
h
ay
h
up
p
l
y
7
,A
a
a
l
e
k
k
e
ex
x
)
)
7
7
3
34
4|
|3
31
5|
18
|
3
86
3
6
S
2
2
0
SY
05
5
t
Y
N
5A
th
N
AC
ha
,
AC
C
K
C
K
31
1
8
8
6
6|
|7
7
3
35
5
A
A
C
C
K
K
3
3
1
18
8
6
6|
5|
|7
7
|7
3
35
73
5s
35
so
5m
o
m
mo
me
o
r
r
e
ed
e
d
da
da
at
at
ta
t
a
a
3
a
3
1
19
95
a
t
t
r
ra
a
n
ns
d
d
a
sm
at
m
i
ta
is
s
s
s
i
io
o
n
n
7
73
35
73
3
5
5|
|
3
32
20
0
5
an
n
k
k
s
s
!
!
3
74
42
2A
AC
C
K
K
3
32
20
05
5
|
|
7
74
4
2
K
h
h
a
an
4
4
–
-w
nd
wa
ds
s
h
ay
y
7
ha
ak
ke
e
7
7
4
42
2|
|
3
3
2
20
06
6A
A
C
CK
K
74
42
2|
|3
3
2
2
0
06
6F
FI
I
N
N
FIT1047
48
TCP parameters
TCP implements segmentation
• large application layer messages are split into segments
How fast to send segments?
• Sending too many at once may overload receiver or intermediate path with lower bandwidth
How to decide the segment size?
• Sending too large segments requires IP to fragment
• Large segments also increase errors
FIT1047
49
TCP segment size
Two approaches:
• use “reasonable” MTU, accept that IP sometimes needs to fragment
• use Path MTU Discovery (PMTUD):
• send IP packets to destination, asking routers to
never fragment
• if router would have to fragment, it sends back an
error message
• increase packet length until error occurs, then use last known error-free MTU
FIT1047
50
TCP congestion control
How fast to send?
• Receiver transmits its maximum buffer size
• Sender sends segments without waiting for ACK up to buffer size
What if network cannot cope?
• Start slow: wait for ACK after each segment
• Increase with every ACK: send two, four etc segments after each ACK
• Fall back to slower speed when no ACK arrives
FIT1047
51
Layers of Abstraction
Switch Router Switch
FIT1047
52
The Internet
FIT1047
The Internet
• A network of networks
• Connecting millions of networks
and billions of devices
• Based on a common, standard set of protocols
Source: Wikipedia / Opte project
FIT1047
54
Autonomous Systems
Networks operated by a single organization
• e.g. Monash University’s or your ISP’s network
Interior routing
• for routing packets within an AS
• uses RIP, OSPF, EIGRP
Exterior routing
• for routing packets between different AS
• Internet uses BGP (Border Gateway Protocol)
FIT1047
55
Internet Architecture
AS1 OSPF
g
I
In
n
t
t
e
er
ri
i
o
o
r
r
r
ro
ou
ut
t
i
in
n
g
BGP
AS3 BGP
B
Bo
or
rd
d
e
e
r
r
r
r
o
ou
u
t
t
e
er
r
BGP
BGP
AS4
RIP AS2
BGP
BGP
BGP
OSPF
AS5
AS6
FIT1047
56
Internet Structure
Internet Service Providers
• Commercial companies
• You connect via ADSL, 4G, cable, NBN, …
• How are they connected with each other?
Hierarchy of ASs
• Each ISP operates an AS
• Routing information shared between ASs using BGP
• ISPs connect at IXPs
FIT9135
57
Internet Structure
r
r
o
l
o
la
a
u
ut
r
rg
te
ge
e
s
er
s
v
rI
IS
S
P
S
vi
ia
P
P
P
a
E
Ex
xc
c
h
I
I
n
nt
ha
a
t
e
er
n
n
c
r
n
c
n
e
e
e
eP
t
t
Po
o
i
i
n
n
t
t
s
sm
m
a
a
l
ll
l
e
e
r
r
I
IS
P
Pe
P
s
P
e
e
e
er
ee
rs
er
r
s
s
FIT9135
58
Peering
Tier-1 ISPs
• Large ISPs with large WANs
• Charge smaller ISPs for routing their traffic
Peering agreement between two ISPs
• accept each other’s traffic without charge
• usually because both are similar size (similar amount of traffic)
• connect at an IXP
IXP
• provides the hardware for several ISPs to connect
• often owned by a consortium of ISPs
FIT9135
59
Inside an IXP
https://www.convergedigest.com/search/label/DE-CIX
FIT9135
60
Q: IXP devices
What kind of network devices are used to build IXPs?
1. Hubs, all peers need to be able to talk to each other
2. Switches, peers need a layer-2 connection
3. Routers, the traffic must be routed from one ISP to another
4. Firewalls, to ensure only legitimate traffic is exchanged FIT1047
61
Delivering Content over the Internet
FIT1047
The problem
The Internet has grown
FIT1047
63
The problem
The Internet has grown
• not only the number of hosts, but also their distance
• many applications rely on low latency (video, streaming music, modern web sites)
• some services have millions of users
• the main protocols (HTTP, TCP/IP and BGP) were not designed for this growth
So why does it still work?
• Load balancing
• Content Delivery Networks (CDNs)
FIT1047
64
Load Balancing
Many services impossible for single server
• Google processes over 40,000 search queries per second
• 6000 tweets are sent per second
• Netflix streams around 10.2 Tbps on average
Spread load over multiple servers
• DNS-based: host name maps to multiple IPs
• Special hardware: load balancer accepts requests, routes them to different servers
FIT1047
65
DNS-based load balancing
Inside Monash network:
• PING www.google.com (216.58.220.132)
64 bytes from 216.58.220.132: time=13.752 ms
From Optus network:
• PING www.google.com (74.125.237.209)
64 bytes from 74.125.237.209: time=52.074 ms
From Germany:
• PING www.google.com (173.194.112.176)
64 bytes from 173.194.112.176: time=1.43 ms
From France:
• PING www.google.com (74.125.21.105) 64 bytes from 74.125.21.105: time=104 ms
FIT1047
66
Content Caching
Store web data closer to users
• replicate web pages etc. in caches
• can be implemented transparently:
• user makes request
• router on path to server queries cache engine
• if content available, serve from local cache
This is explicitly supported by HTTP
• GET requests can be cached
• HTTP headers can contain Expires: field
• Cache only serves GET requests that are not expired
FIT1047
67
Content Delivery Networks (CDNs)
Load balancing only solves half the problem
• once the requests arrive in your network, you can distribute them to all your servers
• but the requests and responses need to be routed through the Internet
CDNs
• operate servers in multiple locations
• operate their own high-bandwidth network
• locate points of presence close to end users
FIT1047
68
Q: CDNs (Content Delivery Network)
How much of the Internet’s traffic, do you think, is delivered through CDNs?
1. 10% 2. 20% 3. 30% 4. 50% 5. 70%
http://blog.streamingmedia.com/wp-content/uploads/2013/07/2013CDNSummit-B102A.pdf
Report according to deepfield.com
FIT1047
69
CDNs and Peering
Get close to your customers
• improves user experience (fast page load times)
• network inefficiencies are not blamed on you
CDNs are present at IXPs
• peer with anybody for free
• small ISPs avoid paying e.g. for YouTube content downloaded from the upstream ISP
• Example: Netflix peers with Australian ISPs, which can offer “unmetered” access
FIT1047
70
Summary (I)
Network layer
• subnet mask identifies other devices in same subnet
• routers use routing tables to determine next hop
• static vs dynamic routing
• IPv4 vs IPv6
Transport layer
• splits up long messages into segments
• uses ARQ for error control
• uses ports to identify applications
FIT1047
71
Summary (II)
The Internet
• consists of Autonomous Systems
• connected by routers that use BGP for sharing routes
• ISPs run autonomous systems, connect at IXP
• Load balancing and Content Delivery Networks enabled the massive growth
FIT1047
72