Transcript Addressing - University of Windsor

Addressing
The ‘What’ and ‘Where’ of
Communication
Addressing

Addressing is necessary for any
communication
–
–
–
–
–
–

To talk: Appearance, name, …
To call: Telephone numbers
To mail: Postal address
To visit: Postal address + directions
To E-Mail: E-Mail addresses
To instant message: ICQ#, AIM ID, etc.
These ‘addresses’ allow us to uniquely
identify the entity with which we wish to
communicate
Addressing a la Shoch

Name/Identifier: What
– Names normally identify the entity
– If an entity moves, the name/identity will remain
the same

Address: Where
– Addresses identify the location of the entity
– If an entity moves, the address will change

Route: How to get there
– Routes identify the path to get to an entity
– If an entity moves, the route will change
Addressing
Addressing deals with how to define an
entity’s location (uniquely)
 Addressing is necessary for message
delivery

– An address is the start and end point for
the route
• However, routing is another subject
– Where do we want the message to go?
Addresses

We have already seen MAC addresses (for
Ethernet and some other LANs):
–
–
–
–

e.g. 02-60-8C-08-E1-0C
6 octet address
Globally unique
Defined statically by the hardware manufacturer
Most people are familiar with the IP
addresses used by TCP/IP networks:
–
–
–
–
e.g. 137.207.32.2
4 octet address
Not necessarily globally unique
Defined dynamically by DHCP servers or
negotiated by the operating system
IP Addressing
A Closer Look
IP Addresses

TCP/IP networks use IP for the network layer
protocol
 IP defines 4 octet addresses
– 4 billion possible addresses

Usually written in the form A.B.C.D
– A, B, C, and D are each 1 octet (0-255), normally
written in decimal notation
– Thus, IP addresses fall in the range:
0.0.0.0 – 255.255.255.255
IP Addresses

Originally intended for separate
internets (interconnected LANs)
– Thus, the 32 bit size was not a concern
– 48 bits is generally considered a fairly safe
size for globally unique addressing
– Computers connected to ARPANET (and
later incarnations) were just given
consecutive addresses
1.0.0.0, 1.0.0.1, 1.0.0.2, …
IP Addresses
Any computer connected to a TCP/IP
network (e.g. the Internet) must have an
IP address
 Further, any network interface card
(NIC) using TCP/IP to access an
network (e.g. the Internet) must have a
different IP address

IP Addresses
Even though there are 4 billion possible
IP addresses, they are running out
 Here’s why:

– Some of the bits are dedicated to header
information (discussed later)
• ½ the addresses for each lost bit
– Addresses are categorized, and some of
the categories are running out of
addresses (while others are not)
Non-Classed Addresses

Part of the address represented the network
the computer resided on, and part
represented the computer itself
– Network: 7 bits (up to 128 networks)
– Computer: 24 bits (up to 1.6 million computers on
each network)

Since there were very few networks on
ARPANET originally, this wasn’t a problem
Address Classes

When private organizations started
joining the Internet, the needs became
obvious
– Some (fewer) networks have multitudes of
computers (thousands)
• e.g. The @Home network
– Some (many) networks have very few
computers (a few hundred or less)
• e.g. The Windsor Police Department
Address Classes

Quickly, the addresses were separated
into 3 classes (plus room for more
classes if needed):
– Class A: Fewer networks, many nodes
– Class B: Medium networks, medium
nodes
– Class C: Many networks, fewer nodes
IP Address Classes
Class A:
bit index: 0
1-7
0 network
Class B:
bit index: 0 1
1 0
2-15
network
Class C:
bit index: 0 1 2
1 1 0
3-23
network
8-31
host (machine)
16-31
host
24-31
host
IP Address Classes

Class A:
– Range: 1.0.0.0 – 126.0.0.0
– Networks: 128 max, Machines: 65537-1.6 million
– e.g. huge networks, such as large
military/government organizations (e.g. FBI), the
@Home network, etc…

Class B:
– Range: 128.1.0.0 – 191.255.0.0
– Networks: 16384 max, Machines: 257-65536
– e.g. Internet service providers (ISPs) (dial-up)

Class C:
– Range: 192.1.0.0 – 223.255.255.0
– Networks: 2 million max, Machines: 1-256
– e.g. Small businesses
IP Address Classes

The IP address classes are self-identifying
– Which means that given the address, you can
determine what class an address is
• Actually, using only the first number
– Examples:
• 137.207.32.2 (server.uwindsor.ca)
– 137 -> Class B
• 24.0.0.1 (@Home DHCP server)
– 24 -> Class A
Other IP Address Classes
Class D:
bit index: 0 1 2 3
1 1 1 0
4-31
Multicast group address
•These addresses are used to represent multicast groups
•Discussed later
Class E:
bit index: 0 1 2 3 4
1 1 1 1 0
5-31
Reserved for future use
•These addresses were left open to be used and divided
into classes as needed
Special IP Addresses

0.0.0.0: Used to indicate that this machine is
without an assigned IP
– Used during bootstrapping (e.g. requesting an IP
from a DHCP server)

<all 0s (binary)><hostID>: Used to send
messages to some machine on this network
 255.255.255.255: Used to send broadcast
messages across this machine’s network
 <netID><all 1s (binary)>: Used to send
broadcast messages to the specified network
 127.0.0.1: Used to send messages back to
this machine (called loopback or localhost)
IP Addressing Comments

In IP addressing:
– 0’s usually represent ‘this’
– 1’s usually represent ‘all’

Broadcasting, although discussed here
in terms of addressing, will be
discussed further
Loopback

The 127.0.0.1 address, does not normally
exist on the network
– Either as the source address or destination
address of a packet

The address is used internally by NICs
– When a NIC receives a message addressed with
127.0.0.1 to be transmitted, it passes the message
directly to the receiver hardware
– The receiver hardware returns the message to the
operating system exactly as if the message were
received from the network
• However, the message never entered the network
medium
Internal IP Addresses

Depending on the address class needed by
an organization, a range of internal
addresses is available:
– Class A: 10.0.0.0 – 10.255.255.255
– Class B: 172.16.0.0 – 172.31.255.255
– Class C: 192.168.0.0 – 192.168.255.255

IP routers outside a private (connectionshared) network, will not forward datagrams
designated for addresses in these ranges
Multi-homed Machines

There is no restriction preventing
machines from participating in multiple
networks
– A machine could have multiple NICs
– Each NIC would have its own MAC
address
– On TCP/IP networks, each of these NICs
would be given a different IP address
Multi-homed Machines
192.168.0.1
192.168.0.2
192.168.0.3
M
M
M
M
192.168.0.4M
192.168.0.8
Class C private network
192.168.0.7M
M
172.16.3.17M
M
M
M
192.168.0.6
192.168.0.5
172.16.3.16
M
M
M
M
172.16.3.15
172.16.3.14
M
172.16.3.18
Class B private network
172.16.0.1
172.16.0.2
172.16.0.3
172.16.0.4
Multi-homed Machines
192.168.0.1
192.168.0.2
192.168.0.3
192.168.0.8
192.168.0.4
Class C private network
192.168.0.7
172.16.3.17
192.168.0.6
172.16.3.16
192.168.0.5
172.16.3.15
172.16.3.14
172.16.3.18
Class B private network
172.16.0.1
172.16.0.2
172.16.0.3
172.16.0.4
Routers

Routers are multi-homed machines
– They have a number of network ports, each of
which represents a different path

Routers use tables that relate destinations to
network paths
– Internet routers relate destination network
addresses with one of their network ports
– When a datagram arrives at a router:
• Its destination address is used to determine the network
address
• The network address is used to look up the destination
port in the routing table
Network Addresses

An IP address can be used to calculate the
address of the network
 The machine address is passed through a
filter (called a subnet filter):
– This filter extracts the bits of the address that
represent the network and sets the bits that
represent the machine to zero
– The filter determines which part of the address
represent the network address, by using the
subnet mask
Subnet Mask

The subnet mask is a binary number, that has
0s in the machine portion of the address, and
1s in the network portion
 Most networks of each type use a constant
subnet mask
– Class A: 255.0.0.0
(Binary: 11111111000000000000000000000000)
– Class B: 255.255.0.0
(Binary: 11111111111111110000000000000000)
– Class C: 255.255.255.0
(Binary: 11111111111111111111111100000000)
Using Subnet Masks

Example:
– Address: 137.207.32.2
– Subnet Mask: 255.255.0.0
Address:
10001001110011110010000000000010
Mask:
11111111111111110000000000000000
Net Address: 10001001110011110000000000000000

Network address: 137.207.0.0
Routing in Action
Internet
Network N1 (Class B)
Address: 137.207.0.0
P4
P1
137.207.0.0
P1
194.201.61.0 P2
P2
24.0.0.0
P3
*
P4
Network N2 (Class C)
Address: 194.201.61.0
P3
Network N3 (Class A)
Address: 24.0.0.0
IPv6
Next Generation Addressing
in TCP/IP Networks
IPv6
Due to the limited nature of existing IP
addressing (IPv4), a new version of IP
addressing was developed
 This new scheme uses 16 octets for
addresses, instead of 4 octets
 Written using hex notation:

3A57:0000:0000:9CD5:3412:912D:6738:1928
IPv6 Features

16 octet addresses (128 bits)
 Larger numbers of address classes
– More accurate control of network/machine counts

Variable-sized headers
– Optional information can be placed into the header
when needed
– Reduces header size in most cases

Extendible protocol
– IPv6 allows for new header information to be
added to support different protocols
IPv6 Features

Automatically reconfigurable
– Addresses can be automatically reassigned
dynamically
– e.g. when a certain number of nodes join the
network, a different address class may be desired

Autoconfigurable
– The use of autoconfiguration (such as DHCP)
allows dynamic private addressing and dynamic
public addressing
IPv6 Datagram Format
optional
header
extension headers
data
IPv6 Header Format
0
4
version
12
31
traffic class
32
flow label
48
payload length
64
56
next header
63
hop limit
96
source address
128
destination address
IPv6 Integration

Will IPv6 replace IP addresses?
– Who knows?

Currently, temporary solutions have made
IPv4 addresses capable of lasting longer than
originally predicted
 If and when IPv6 is to be integrated, the
process must be a transition
– Closing the entire Internet down to convert
hardware and software to IPv6 not going to
happen
– Some stations may take longer to transition than
other stations
• e.g. Bob’s Internet Shack vs. the Telus Network
IPv6 Integration


NAT (network address translators) provide one
example of such a temporary solution
NATs provide three benefits:
1.
NATs provide IP masquerading
•
2.
Messages using these addresses pass through a network
address translator (NAT) to be transformed into external IPs
NATs provide IP sharing
•
ISPs for example, have many customers, but significantly
less at any given time are logged onto their system
–
3.
IP addresses can be assigned dynamically to these customers
when they log in
NATs provide schemes to allow networks to use either
IPv4 or IPv6
–
Addresses would be converted as they pass through a NAT
IPv6 Integration

Another method that may be used for the
transition between IPv4 and IPv6 is address
inclusion:
–
IPv4 addresses could be embedded into IPv6
addresses
•
–
Translation between the two types of addresses is
possible without any other information
Some problems exist with this approach, but in
general it simplifies communication between
IPv6 networks and IPv4
Special IPv6 Addresses

0:0:0:0:0:0:0:0 Used to indicate that this
machine is without an assigned IP
– Used during bootstrapping (e.g. requesting an IP
from a DHCP server)

0:0:0:0:0:0:0:1 Used to send messages back
to this machine (called loopback)
– These two addresses are not valid on the actual
network medium (same as with IPv4)
00:… Reserved (including IPv4 and IPX
address inclusion)
 FF:… Multicast addresses

IPX
Internetwork Packet
Exchange Addresses
IPX
IPX was originally created to replace IP
 In reality, it is used primarily on LANs
 In conjunction with the SPX protocol,
formed one of the two protocol suites
used in Netware networks

– SPX is to TCP, what IPX is to IP

Still can be (although rarely is) used
today in Windows networks
IPX Addresses

IPX uses a 2 component address (like
IP):
– The network portion (4 octets)
– The machine portion (6 octets)

Unlike IP, these sizes are constant
– So there are no IPX address classes

IPX uses sizes large enough to
accommodate all categories of networks
IPX Addresses

The network portion of an IPX address is 4
octets (32 bits)
– This allows for 4.29*109 networks (4 billion)
– This is almost enough for everyone on earth to
have their own network

The machine portion of an IPX address is 6
octets (48 bits)
– This allows for 2.81*1014 machines on each
network (281,475 trillion)
IPX Addresses
The scalability of IPX addresses is not
their only benefit
 The constant size of the network and
machine address portions simplifies
extracting each portion

– As a result, machines that process IPX
networks can process IPX datagrams more
quickly
• Such as network nodes, routers, etc.
IPX Addresses

Why 48 bits for the machine portion?
– 48 bits allows for way too many machines, more
than will be needed for many years
• By the time machine IDs run out, network hardware and
software will have been obsolete by many years!
– Using 48 bits allows hardware to use the
machine’s MAC address as the machine portion
• This makes auto-configuration (dynamic IPX address
assignment) easier/faster
IPX Addresses

If IPX has been around since the NetWare
days, why don’t we use it for the Internet,
instead of IP?
– Good question! Why don’t we?
– Frankly, IPX has a bad reputation, because initially
it was used with SPX and other bandwidthmunching protocols
– IPX can be used in conjunction with TCP
(TCP/IPX), and it would make an excellent
replacement for TCP/IP
• However, standards organizations (e.g. ISO) want to use
the protocols they develop, and not ones developed by
corporations, such as Novell
Fragmentation & Reassembly

Packets can arrive out of order in
connectionless networks
 Packets must be reordered during
reassembly
 During fragmentation, the portion of data that
each fragment represents must be identified
– Since the length of a packet’s data can always be
determined, all that is necessary is to use the
offset of the start of the packet’s data in the larger
data chunk
Fragmentation & Reassembly
0
1500
3000
4500
6000 6800
Logical Data Chunk
Length:1500 Length:1500 Length:1500 Length:1500 Length:800
Offset:0
Offset:1500 Offset:3000 Offset:4500 Offset:6000
Packet 1
Packet 2
Packet 3
Packet 4
P5
Fragmentation & Reassembly

Why do we use the data offset, and not just a
sequence of numbers to determine packet
order?
– Sometimes, packets can be fragmented at one
location, and must be re-fragmented at another
location (such as while passing through a network
incompatible with larger frame sizes)
– These situations would require renumbering of all
packets in the sequence, which is not always
possible
Fragmentation & Reassembly
Length:1500 Length:1500 Length:1500 Length:1500 Length:800
Offset:0
Offset:1500 Offset:3000 Offset:4500 Offset:6000
Packet 1
Packet 2
Packet 3
Packet 4
P5
Packet 1
Packet 2
P3a
Packet 4
P5
Length:1000
Offset:3000
P3b
Length:500
Offset:4000
Fragmentation & Reassembly

Re-fragmentation (at gateways, routers, …) is
expensive
– The re-fragmenting node must process each
packet, fragmenting it into smaller packets
– Another reassembling node must collect these
packets and assemble them into larger packets
– Each of these operations involved memory
processing, which is expensive when applied to
many packets per second
Fragmentation & Reassembly

Re-fragmentation (particularly in routers)
should be avoided at all costs
– To virtually eliminate re-fragmentation in a
network, the maximum transmission unit (MTU)
should be determined and used as the packet size
– A network’s MTU is the largest size that can be
used for packets that will not result in any refragmentation by any routers, or other multihomed nodes
– Schemes for determining the MTU dynamically
have been developed, but are beyond the scope
of this course
IP Datagrams

IP datagrams are packets sent over IP
networks using connectionless messaging
 Datagrams can be used directly within
network-capable programs by sending
datagrams via UDP (user datagram protocol)
 Datagrams are used transparently by TCP to
provide connection-based transport
IP Datagrams
bits
Name
Description
Used For
4
Version
Version (equal to 4)
All
4
IHL
Header length
All
8
TOS
Type of Service (obsolete)
-
16
Length
Total length of datagram (header included)
All
16
ID
Identifier: used in reassembly to identify packets
Reassembly
1
DF
Should the datagram be re-fragmented, if necessary?
Routers (re-frag)
1
MF
Are there more fragments in the sequence?
Reassembly
13
Offset
Offset of data that this datagram represents
Reassembly
8
TTL
Hop limit
Routers
8
Protocol
Transport protocol used for this packet (UDP, TCP)
Acknowledgement
16
Checksum
Checksum of the header
All
32
SA
Source address
All
32
DA
Destination address
All
?
Options
Future features
-
?
Padding
Fills remaining space
-
IP Datagram Routing

When an IP-enabled router receives a datagram, it:
– Receives a datagram through one of its ports
– Deletes the datagram, if the hop count (TTL in IPv4, Hop limit
in IPv6) has a non-positive value
– If the hop count is positive, it is decremented and processing
continues
– Determine the destination address’ network address
– Uses the destination network address to find an entry in the
routing table
– Uses the routing table entry to determine to which port the
datagram should be sent
– Sends the datagram through the correct port
IPv6 Datagrams
bits
Name
Description
Used For
4
Version
Version (equal to 6)
All
8
TOS
Type of Service (status info)
All
20
Flow label
Future features
-
16
Length
Length of data in the datagram (header not included)
All
8
Hop limit
Hop limit (decremented to zero)
Routers
16
SA
Source address
All
16
DA
Destination address
All
?
H2H
Hop to hop header
Routing
?
SRH
Source routing header
Routing
?
FH
Fragment header
Reassembly
?
E2E
End to end options
Reassembly
Header Checksums



Networks sometimes result in corrupt data
Information in the header is equally susceptible to this
corruption
However, header information, when corrupt, can cause
more serious difficulties
– For example, the destination address may have a few bits
changed, or the hop count, etc.
– Corruption like this, is not always easy to detect and fix
– Corrupt data (determined by another checksum) can be fixed
by re-issuing the datagram
– Header checksums are used to ease identification of header
corruption