IPv4 vs.

IPv6: Much of the information within this tool is intended for use with IP version 4 (IPv4), the current IP standard.
IP version 6 (IPv6) is on the horizon, however, and you should understand the differences between it and IPv4. To help you
do so, we've included a section comparing IPv6 and IPv4 on page 4.

TCP/IP DEFINED

Transmission Control Protocol/Internet Protocol (TCP/IP) is an industry-standard suite of protocols that computers use to find, access, and
communicate with each other over a transmission medium. A protocol is a set of standards and rules that need to be followed. In the case of
networking computers, a protocol is the set of standards and rules that a machine's hardware and software must follow in order to be recognized
and understood by other computers. The protocol suite is implemented via a software package most commonly known as the TCP/IP stack. There
are four general layers of the TCP/IP stack:
Application layer Transport layer Internet layer Physical or Network Interface layer

The TCP/IP reference model and OSI reference model

The Open Systems Interconnected reference model (OSI/RM) is the standard that all other protocols follow. The OSI/RM provides a framework that
connects heterogeneous systems using a common protocol. It also gives developers universal concepts so they can develop and perfect protocols.
© 1995-2004 CNET Networks, Inc. All rights reserved. “TechRepublic” and its logo are trademarks of CNET Networks, Inc. Reproduction of this publication in any form is prohibited.

Each layer of the TCP/IP reference model corresponds to a part of the OSI model.

OSI TCP/IP
Sender Reference Reference Recipient
Model Model

Application layer: Contains network applications

Application Windows Sockets
Presentation
Session
TDI Interface
TCP
Connection-oriented
protocol that establishes a con-
Transport
nection with another machine
and maintains that connection for
the duration of data transmission.
More reliable and more widely
used but slower than UDP.
and services that the user interfaces with in order to
NetBIOS NetBIOS
use network communication. Contains utilities for things
over like file and print services and name resolution (for
TCP/IP example, NetBIOS, an application programming inter-
face [API] that supports a desktop operating environ-
ment). Contains utilities that work with TCP/IP to pro-
vide connectivity, file transfer capabilities, utilities for
remote administration, and Internet utilities (for exam-
ple Ping, TRACERT, FTP, and Telnet).
UDP Transport layer: Contains the Transmission
Connectionless protocol that Control Protocol (TCP) and the User Datagram
does not establish a persistent Protocol (UDP). Used for network connectivity by
applications. Ensures that data is received by the right
connection with the target
machine and that the right application is running on
machine. Faster but less
that machine. Ensures the integrity and completeness
reliable than TCP. of transmitted data via error checking, flow control,
and verification.

Internet layer: Contains Internet Protocol (IP),

DATA

DATA
Address Resolution Protocol (ARP), Internet Control
ICMP ARP Message Protocol (ICMP), Reverse Address
IGMP IP
Network RARP Resolution Protocol (RARP), and Internet Group
Management Protocol (IGMP). IP addressing and
address resolution occur within the Internet layer. ARP
resolves logical IP addresses into physical equivalent
addresses. ICMP is mostly used by routers to send
NDIS Interface information back to a source computer about a trans-
mission that computer is trying to make. When you
use the Ping utility, the information you receive was
gathered using ICMP.

NIC Drivers Network Interface layer: Contains a collection

of services and specifications that provide and man-
Data Link age access to the network hardware. Its responsibili-
Ethernet PPP ties include:
Physical
FDDI Frame Relay • Interfacing with the computer's network hardware.
• Checking for errors in incoming packets of data.
• Tagging outgoing packets with error-checking
Network Adapters information.
• Acknowledging the receipt of a packet.
• Resending that packet if no acknowledgment is
returned by the recipient.

The layers of the OSI model and how they map to different areas of Microsoft’s TCP/IP, and the four layers of the
TCP/IP reference model and how they map to Microsoft’s TCP/IP

This is a single-user version, and duplication is prohibited. For additional copies, please call 1-800-217-4339.
 IP ADDRESSING

The TCP/IP protocol suite uses an Internet address, or IP address, to
uniquely identify each host or router on an internetwork. What makes IP
addresses unique is that each address identifies only one host or router
on the Internet. Any host or router that wants to be connected to the
Internet must use the addressing scheme used by TCP/IP.
An IP address is a 32-bit (4-byte) number written in dot notation,
which simply means that each byte is separated by a dot, or period. A
typical IP address in dot notation would look like this: 192.168.10.21.
The same address written in binary format would be:
11000000.10101000.00001010.00010101
Each byte in an IP address can represent any number from 0 to 255.
With 4 bytes used in each IP address, this means that the total number
of IP addresses available is 232, or 4,294,967,296 possible IP
addresses. Every IP address contains two distinct parts:
The netid identifies the network.
The hostid identifies the host on that network.
IP address classes
Although a lot of IP addresses are available, you don't just pick out any
address for use with your company network and assign addresses
© 1995-2004 CNET Networks, Inc. All rights reserved. “TechRepublic” and its logo are trademarks of CNET Networks, Inc. Reproduction of this publication in any form is prohibited.
reserved for special use. This means that the actual maximum number
of hostid addresses on a class A network is 224 minus 2, or
16,777,214. This limitation applies to all IP address classes.
Class B addresses: Assigned to organizations with large networks.
Class B networks have the following characteristics:
Class B network addresses range from 128.0.0.0 to
191.255.255.255.
The first 2 bits in a class B address are always 10 (128).
In a class B network, the netid is determined by the first 16 bits
(first 2 bytes).
There are 214, or 16,384, possible class B networks.
The hostid is determined by the last 16 bits (last 2 bytes). Because
an address with all bits set to 1 or all bits set to 0 is not allowed, the
maximum number of hostids on a class B network is 216 minus 2, or
65,534 hostids.
Class C addresses: Assigned to organizations with small to medium
networks. The class C network is the most common network in use
today. Class C networks have the following characteristics:
Class C network addresses range from 192.0.0.0 to
223.255.255.255.

however you like. Depending on your networking requirements, you are Class C networks use the first 24 bits to determine the netid.
assigned (by your service provider) a specific class of IP address. The three leftmost bits in a class C network are always 110 (192).
There are five classes of IP addresses: A, B, C, D, and E. Table A The next 21 bits are used to define the network.
shows you the most commonly used classes. There are 221, or 2,097,152, possible class C networks.
Class A addresses: Assigned to organizations with very large In class C networks, 8 bits are used to define the hostid. Because
numbers of computers—including servers and routers—attached to an address with all bits set to 1 or all bits set to 0 is not allowed, the
their networks. A government department or a Fortune 500 corporation maximum number of hostids on a class C network is 28 minus 2, or
would be examples of this type of network. The following characteristics 254 hostids.
apply to class A networks: Class D addresses: Reserved for multicasting. Class D networks
Class A addresses range from 0.0.0.0 to 127.255.255.255. have the following characteristics:
The leftmost bit in a class A network is always 0. In class D addresses, the first 4 bits are always 1110 (224).
The first 8 bits (first byte) define the netid. The remaining 28 bits are used to define multicast addresses.
The remaining 24 bits define the hostid. There is no netid or hostid in a class D address. The whole address
It would appear that because the first 8 bits define the netid, the is used for multicasting.
maximum number of class A networks available would be 27, or 128 Class E IP addresses: Reserved for special use on the Internet.
networks. However, a netid with all bits set to 1 and a netid with all bits There is no netid or hostid in a class E address. The first 4 bits in a
set to 0 are reserved for special use. This means that the actual class address are always 1111 (240).
number of class A networks available is 27 minus 2, or 126 networks.
Table B lists the number of netids and hostids available for each IP
address class.
Class A addresses use 24 bits to define the hostid. In theory, there is
a maximum of 224, or 16,777,216, hostids available. Like netids, a
Special IP addresses: Class A, B, and C addresses each have
addresses set aside for special purposes.
hostid with all bits set to 1 and a hostid with all bits set to 0 are
Network address: In class
A, B, and C addresses, any
address with a hostid with all
Table A: Commonly used IPv4 IP address classes bits set to 0 is used to define
the network address. This
Class First Octet Subnet address is never assigned to
Decimal Hex Binary a host. This simply means
that your network itself is
Class A 1-127 255.0.0.0 FF.0.0.0 11111111.00000000.00000000.00000000
given an IP address with all
Class B 128-191 255.255.0.0 FF.FF.0.0 11111111.11111111.00000000.00000000 the hostid bits set to 0. This
Class C 192-233 255.255.255.0 FF.FF.FF.0 11111111.11111111.11111111.00000000 is not the same as the netid.
The network address is
always an address with the
Table B: Number of netids and hostids host bits set to 0. For example, if a host on a network has an IP
available for each IP address class address of 142.23.120.24, the network address is 142.23.0.0.
This host on this network: If an IP address consists of all zeros, the
Class Number of Number of address refers to this host on this network. A host uses this address
network addresses host addresses when it is booted, but the host does not know its IP address. This
7 24 address is always a class A address, regardless of the network
Class A 2 -2 (126) 2 -2 (16,777,214)
configuration.
Class B 214 (16,384) 216-2 (65,534)
Specific host on this network: An IP address with all netid bits set
Class C 221(2,097,152) 28-2 (254) to 0 refers to a specific host on this network. Setting the netid bits to 0
Class D None None ensures that any IP information sent using this address is not routed.
The information will remain within the local network. For example, if my
Class E None None

This is a single-user version, and duplication is prohibited. For additional copies, please call 1-800-217-4339.
 IP ADDRESSING continued

IP address is 216.120.46.100 and the machine is on the 216.120.46
local network, it could send information to a host with the IP address
216.120.46.69 by using the address 0.0.0.69.
Direct broadcast address: Any class A, B, or C address with the
hostid set to all ones is known as a direct broadcast address. Routers to
send information to all hosts on a specific network use this address. For
example, if you sent IP information to 204.36.120.255, all hosts on the
204.36.120.0 network would receive the information sent.
Limited broadcast address: Any class A, B, or C address with all bits
set to 1 (255.255.255.255) is used to send information to all other hosts
on the same network as the host sending the information. In other words,
this address sends IP information to all hosts on your local network.
Loopback address: Any IP address with the first byte set to 127 is
used for the loopback address. The loopback address is used to test the
TCP/IP software on a computer, and it does not provide any information
on the configuration of a network interface. The most common loopback
address is 127.0.0.1. When you run the command ping 127.0.0.1, you
can determine whether your TCP/IP software is running properly. A child
process on a computer may also use the loopback address to send a
message to its parent process.
Private IP networks: There are three blocks of IP addresses that may
© 1995-2004 CNET Networks, Inc. All rights reserved. “TechRepublic” and its logo are trademarks of CNET Networks, Inc. Reproduction of this publication in any form is prohibited.
last octet is not 0 (however, the leading octets of the supernet mask are
still 255). Supernetted addresses will require a default gateway that
needs to be supernetted as well.
Address ranges, or blocks, are important in supernetting. They allow
you to identify the valid addresses in a tabular format that helps identify
boundaries on networks. There are many tables you can create or find
on the Internet to plan your networks when using supernetting. Table D
shows a supernetting chart using an example configuration.
This is a chart of the /17 through the /30 block of class C supernets.
These ranges are scalable, helping you select how many networks and
hosts you would like to use. You may notice that /24 CIDR block looks
familiar, as that is really not a supernetted network but a subnetted
single class C network with a standard 24-bit subnet.
Calculating a supernet address is easy if the approach is organized.
Using Table D, determine how many hosts you want to have available on
your network and reference that against the # of Hosts column to select
the best match. Then, once you select the appropriate number of hosts,
you can look across the chart and see the corresponding supernet
mask. With that, you will need to determine a valid starting network. This
starting network must meet certain criteria:
All networks are consecutive from your starting network.

be used by any organization for TCP/IP networking. These IP addresses The third octet of the first network must be an even number (zero is
are nonroutable, meaning they are limited to use within the local network valid for certain situations).
only. The three groups of nonroutable IP addresses are listed in Table C.
When combining eight networks (like the example below), the
third octet of the network number must be evenly divisible
Table C: Private addresses (nonroutable) by eight.
Class Start End Subnet Create a table listing the available networks, addresses, supernet
mask(s), default gateway(s), and other networking objects to outline
Class A 10.0.0.0 10.255.255.255 10.255.255.255 the network.
Class B 172.16.0.0 172.31.255.255 172.X.255.255
Class C 192.168.0.0 192.168.255.255 192.168.X.255
Plan before subnetting or supernetting
Plan before subnetting or supernetting a network. Find out how many
host addresses will be needed and how much room will be needed for
Subnetting and supernetting expansion. It's easier to make room when first subnetting and
supernetting than it is to go back later and resegment a large network.
Originally, networks were divided using the class structure defined by
IPv4. Unfortunately, there just isn't enough middle ground between B
and C. While a standard class C network contains 254 addresses (too
few addresses for a moderate-size company), a class B network has
65,534 (far too many for the average network). This is where subnetting Table D: Supernetting class C addresses
and supernetting can help.
Subnetting is the process of dividing an IP address range into two or This represents part of the CIDR/supernetting chart to help
more separate ranges, called subnets. Breaking networks into smaller, determine which supernet option to choose.
easily defined subnets allows the administrator to better address data CIDR Supernet # of # of
flow and security and creates a logical map that can facilitate Block Mask Networks* Hosts**
troubleshooting. Subnetting involves the use of an IP address and
subnet mask to determine whether a destination network is local or /17 255.255.128.0 128 32,766
remote. /18 255.255.192.0 64 16,382
A subnet mask is a 32-bit number, commonly represented in dotted /19 255.255.224.0 32 8,190
decimal format, that divides the IP address into network and host
/20 255.255.240.0 16 4,094
sections. This allows the sending computer to determine whether routing
will be required when communicating to another system. If the computer /21 255.255.248.0 8 2,046
determines that it is a local address, the packet is sent via Ethernet (or /22 255.255.252.0 4 1,022
another Layer 2 mechanism). If the address is not local, the packet is
/23 255.255.254.0 2 510
forwarded to the default gateway for routing.
Supernetting is the opposite of subnetting. Supernetting is also known /24 255.255.255.0 1 254
as CIDR (classless interdomain routing) as defined by RFCs 1517, /25 255.255.255.128 Less than 1* 126
1518, 1519, and 1520. In IPv4, CIDR is one way of attempting to /26 255.255.255.192 Less than 1* 62
manage the shortage of TCP/IP addresses until IPv6 takes over.
/27 255.255.255.224 Less than 1* 30
Supernetting acts to bridge the gap between a class C network that is
limited to 254 addresses and a class B network that is too large, with /28 255.255.255.240 Less than 1* 14
over 65,000 addresses. In this way, it's possible to have a "logical" /29 255.255..255.248 Less than 1* 6
network that offers the number of hosts that best suits your situation. /30 255.255.255.252 Less than 1* 2
Supernetting achieves this by making a single network that has your
*Number of full class C networks—256 or more available addresses
specified number of hosts and corresponding supernet (like a subnet
mask). A supernetted address will look like any other TCP/IP address in **Available addresses—network and broadcast addresses excluded
dotted decimal format (XXX.XXX.XXX.XXX), but it will have a
supernetted subnet mask. This looks like a normal subnet mask, but the

This is a single-user version, and duplication is prohibited. For additional copies, please call 1-800-217-4339.
 IP ADDRESSING continued

You should also seriously consider using a subnet calculator when
subnetting or supernetting a network. Although you can determine the
correct subnet addresses for your network(s) manually, subnet
calculators can do the math for you with just a few clicks. You can find
many free subnet calculators by entering "subnet calculator" into your
favorite Internet search engine.
While most new routing equipment and current operating systems
support CIDR in their implementation of the TCP/IP protocol, it is critical
to ensure that all components of your network are supernetting-aware
before beginning a supernetting implementation. This includes operating
systems, network services, routers, routing protocols (RIP2, for
example, does not support CIDR), and any network-based services
used on your network.

COMPARING IPv4 AND IPv6

As more and more nodes have joined the
global network, a flaw in IPv4 has
become apparent. For IP addressing to
work, every network device must have a
unique address. When IPv4 was
developed, the 32-bit address space
provided more than enough unique
Denoting IPv6 addresses
addresses. Today, however, the world is
© 1995-2004 CNET Networks, Inc. All rights reserved. “TechRepublic” and its logo are trademarks of CNET Networks, Inc. Reproduction of this publication in any form is prohibited.
Configuration is easier; both stateful
addressing (where addresses are
automatically assigned by a DHCP
server) and stateless addressing (use
of local-link autoconfiguration without
DHCP) are supported.
Anycast addresses, which identify
multiple interfaces but send the packet
only to the nearest interface
Backward compatibility
of IPv6
The designers of IPv6 knew it would not
be practical to have the entire IP world

running out of available IP addresses.
While Network Address Translation (NAT)
partially solved this problem by allowing
private networks to connect a large
number of computers to the Internet
using only one or a few public IP
addresses, NAT is only a stopgap
solution. Internet Protocol version 6
(IPv6), also known as IPng or IP Next
Generation, will be a more permanent
solution.
IPv6 provides for a 128-bit address
space, which will exponentially increase
the number of available public IP
addresses. However, IPv6 offers other
improvements over IPv4:
It supports IPSec for better security
when sending data across a TCP/IP
network.
It supports Quality of Service (QoS) for
better transmission of real-time, high-
bandwidth applications such as video-
conferencing and Voice over IP.
It is more efficient; header overhead is
minimized, and backbone routers
require smaller routing tables.
While IPv4 addresses are traditionally
denoted in decimal format, the longer and
more complex IPv6 addresses are
expressed in hexadecimal format. A
sample IPv6 address looks like this:
21DA:00D3:0000:2F3B:
02AA:00FF:FE28:9C5A
Each hexadecimal number, separated
by colons, represents 16 bits (binary
digits). Zeros at the beginning of a block
can be omitted to simplify the address.
Characteristics of IPv6
addressing
Unlike those on IPv4 networks,
computers on IPv6 networks generally
have more than one IP address assigned
to a single network interface. This is
called logical multihoming.
IPv6 addresses fall into the following
categories:
Unicast addresses, which are used to
identify an individual network interface
Multicast addresses, which identify a
group of network interfaces for
simultaneously sending to multiple
interfaces
move to IPv6 in one massive upgrade.
They ensured that IPv6 would be
backward compatible with v4. As a start,
IPv6 equipment recognizes all v4 packets.
It is also possible to purchase equipment
that encapsulates IPv6 packets in IPv4
packets ("tunneling"). This can be done at
the gateway or router level. So an
organization can roll out IPv6 in two or
more locations and have those locations
communicate via standard v4 networks
and backbones.
The IPv6 stack is also designed to be
run alongside IPv4 stacks. You can "dual-
stack" a machine without a problem. This
means that an end-node or a router can
communicate with both kinds of networks.
It also means that you can upgrade hosts
first, routers first, or any combination that
meets your needs. You can even
configure the IPv6 stack with an IPv4-
compatible address if you want.

ISBN: 1-932509-22-4
Product Code: T021rev1
Price: $12.95 U.S. funds
(Volume discounts available)
Customer Service: 845-566-1866
or toll free 800-217-4339
Visit us online at
TechRepublic.com
Technical Editor: Bill Detwiler © 1995-2004 CNET Networks, Inc.
Technical Contributors: Bruce All rights reserved. “TechRepublic”
Maples, Scott Lowe, Debra and its logo are trademarks of
Shinder, Robert L. Bogue, Thomas CNET Networks, Inc.
Nooning, Rich Vanover, Brien M.
Posey, Greg Shultz, Jim McIntyre
Graphic Artist: Kimberly Wright
Copy Editor: Linda Watkins

Sept 2004

This is a single-user version, and duplication is prohibited. For additional copies, please call 1-800-217-4339.