TCP/IP is a collection of communication protocols that define how different
computers are addressed on the network, what methods are used to move information
from one computer to another, and some services that are available between
computers. The router
primarily deals with the network layer (IP) and the transport layers (UDP
and TCP) in performing its routing and switching functions.
TCP/IP Standards
TCP/IP is often described as an open standard,
which means that no one company or person controls the specifications of the
protocol or the way in which it operates. Instead, a governing body called
the Internet Engineering Task Force (IETF), consisting of networking industry
experts and company representatives, guides the evolution of the protocol.
Working groups within IETF review, discuss, recommend, and approve proposed
changes to the standards by means of vehicles called Request for Comments
(RFC).
All of the concepts and many of the topics covered in this chapter are
defined by the hundreds of RFCs that make up the standards for TCP/IP. Though
often dry reading and technical in nature, the RFCs provide the most complete
definitions of the TCP/IP protocols. RFC documents can be obtained from the Internet Network Information
Center (InterNIC) Web site at http://www.internic.net. As of this writing, Cisco Systems also maintains an archive of RFCs on
their Cisco Connection Online (CCO) Web site at http://www.cisco.com/warp/public/459/
.
IP (Internet Protocol), the addressing component of TCP/IP, operates
at Layer 3 of the OSI model. Each station that wants to communicate
with another has a unique IP address, in the same way that each house on a
street has an individual address. The IP address is a little more complex
than a street address (computers like those 0s and 1s), but after a little
examination, it won't seem so mysterious.
At Layer
4 of the OSI model for TCP/IP are the two primary transport protocols, UDP (User
Datagram Protocol) and TCP. As mentioned previously, the transport protocols
are responsible for basic transfer mechanisms, flow control, reliability,
and error checking of communications between stations. UDP is considered unreliable
because packets
sent using this protocol are not acknowledged by the receiving station. It
is considered connectionless
because a sending station is not required to advise a receiving station of
its desire to form a communications channel over which to pass data. In contrast,
TCP is considered a connection-oriented protocol because
a sending station must advise the receiving station of its desire to form
the communications channel. Packets sent via TCP are tagged with sequence
numbers, and the sending and receiving stations each acknowledge the receipt
of one another's packets.
The IP address is a 32-bit binary address written in four groups of
8 bits called octets. The complete address represents the three
components of the addressing model of IP, namely the network, subnetwork,
and host portions of the address. First, let's look at the address numbering
itself.
A typical IP address written in 32-bit binary might look like the following:
Each of the 8 bits in an octet can take on the value of 0 or 1. Therefore,
the values can range from 00000000 to 11111111 in each
octet. You can already see that managing 32-bit addresses in binary can be
cumbersome and prone to errors. Recognizing that fact, the developers of TCP/IP
decided that the binary should be reserved for computers and that IP addresses
should be converted to decimal form (the common way people look at numbers)
for easier human interaction. Therefore, an octet in which all bit positions
are set to 1 is equivalent to 255 in decimal form:
Adding up the decimal equivalent of the binary number, we
have the following:
Now, let's convert an address example:
Therefore, the decimal notation for this IP address is 172.16.1.1.
The IP address represents the three components of
the IP addressing model: the network component, the subnetwork (often shortened
to subnet) component, and the host component. The three components describe
the different levels of entity specificity within a collection of networked
systems. The host component is most specific, describing
the address of a single workstation or server. The network
component is most general, describing the address of a collection of hosts
within the same logical computer network. The subnet component falls
between the network and host components. It describes the address of a subset
of the hosts within the overall network address space.
The subnet
is created by “borrowing” a portion of the host component to
make address subgroupings within the same logical network. The subnet component
typically identifies a collection of systems within a LAN or WAN segment.
Read from left to right, an IP address moves from the least specific portion
of the address (the network portion) to the next most specific portion (the
subnet) to the most specific portion (the host). Where the breaks between
the three levels occur in the address depends on the address class and how
the address has been subnetted.
As originally specified in the RFCs,
there are five classes of network addresses, which are distinguished
by how many initial bits of the address are set to one:
Originally, Class A network addresses were intended
mainly for very large networks. With Class A addresses, the first bit of the
first octet is reserved and set to 0 and the next seven bits are used to identify
the network component. The three remaining octets make up the host component.
Given these groupings, Class A addresses provide relatively few networks,
but each network can accommodate many hosts within the given address space.
In Class B network addresses, the first two bits of the first
octet are reserved and the first bit is set to 1 with the second bit set to
0. This construction gives Class B addresses 14 bits for the network component
and 16 bits for the host component. Class B network addresses allow for roughly
equal numbers of networks and the hosts on those networks.
In Class C network addresses, the first three bits for the
first octet are reserved and the first two bits are set to 1 and the third
set to 0. This construction gives Class C addresses 22 bits for the network
component and only 8 bits for the host components. There can be millions of
Class C networks; however, each one can support only 255 hosts.
Class D addresses are reserved for multicast
groups. In Class D addresses, the first four bits of the first octet are reserved
and the first three bits are set to 1. A multicast address does not represent
a single station address, but rather a group of stations that want to receive
information. With multicasting, a station can send a single stream of information
to a specific multicast IP address. The network devices—such as routers
and switches—then replicate the stream, sending it to multiple stations
that have registered to receive the data stream.
Class
E addresses are defined by IP. Although they are unused at this time, they
are reserved for future use. In Class E addresses, the first four bits of
the first octet are all set to 1.
Figure 4-1 illustrates the
address structure of network Classes A, B, and C.
When converting an IP address from decimal to binary and determining
how many of the high-order bits are set to one, it is easy to see to which
network class an address belongs. Assuming that there is no subnetting, knowing
the class to which the address belongs tells us what portion of the address
to read as the network portion and what portion to read as the host portion.
Devices such as routers need to decipher this information to deliver data
to the appropriate destination.
If a network is subnetted, however, it is not
possible to tell at a glance how much of the host portion of the address has
been borrowed to make the subnet. To solve this dilemma, IP addresses also
have a subnet
mask (commonly referred to as the network mask).
Like the IP address, the network mask is a 32-bit binary number, grouped into
4 octets, that can be expressed in decimal notation. Unlike the IP address,
however, the network mask has bits set to one in all positions except for
the host portion of the IP address.
For example, a Class B network with no subnetting defined has a mask of
255.255.0.0, in which the upper 16 bits of the mask denote the network portion
of the IP address and the lower 16 bits denote the host portion of the IP
address. A Class B network in which 7 bits of the host portion of the address
have been used for subnetting would have a mask of 255.255.254.0. A Class
C network with 4 bits of subnetting would have a mask of 255.255.255.240.
Figure 4-2 shows
the relationship between the network mask and the IP address.
Subnetting gives network administrators the flexibility
to assign a unique network identifier to each LAN and WAN segment without
having to obtain a separate network address space for each. For example, instead
of a single Class B network address having one logical network segment of
65,000+ hosts, a subnetting scheme that borrows 8 bits from the host component
allows for 255 logical network segments of 255 hosts each. By pairing an IP
address with its network mask, it is possible to determine exactly which bits
of the address correspond
to the network, subnet, and host components. For example, an IP address of
131.108.3.4 with a network mask of 255.255.0.0 has a network component of
131.108.0.0, a host component of 3.4, and no subnet component. An IP address
of 131.108.3.4 with a network mask of 255.255.255.0 has a network component
of 131.108.0.0, a subnet component of 3, and a host component of 4.
With today's routing protocols carrying network mask information as
well as network information in their updates, you can use multiple network
masks within a single logical IP network to increase the efficiency of IP
address utilization.
The concept of the network mask has been extended beyond its original
subnet usage. In response to the explosive growth of the Internet, the number
of IP network addresses requested, the shortage of IP address space, and the
size of the global IP routing table, the agencies that issue IP addresses
might not issue IP addresses along the specified class boundaries described
thus far. Instead, they might choose to group multiple IP network addresses
of a given class together into what is called a supernet, or a
Classless Interdomain Route (CIDR) block.
Additionally, some of the former Class A networks have been subdivided
and issued as smaller CIDR blocks to companies and ISPs. In the past, a company
or ISP might have been issued a Class B network. Today, they might be issued
255 Class C addresses, ranging from 209.32.0.0 to 209.32.255.0. The natural network mask of these Class C networks—with no subnetting within the block of addresses—is 255.255.255.0.
However, by shortening the mask and creating a supernet of these addresses,
the same group of addresses can be represented by the network address 209.32.0.0
and network mask 255.255.0.0. The organization
that receives an allocation of a CIDR block is then free to further subdivide
that network address space as either subnets within their logical network
or as allocations to their customers.
This same method can be applied to the Class A addresses in reverse.
Formerly, the network address 12.0.0.0 with a natural network mask of 255.0.0.0
would have to be assigned to one company or ISP. Now, this network address
can be treated as a block of addresses, and smaller pieces can be allocated
to multiple entities. For example, the group of addresses from 12.1.0.0 through
12.1.255.0 can be represented as a single CIDR block with network address
12.1.0.0 and a network mask of 255.255.0.0. By subdividing these formerly
large network address blocks, larger numbers of IP network
addresses have been made available and address exhaustion has been slowed.
Writing and describing network addresses as four dotted-decimal octets
followed by a four-dotted-decimal octet network mask has always been somewhat
cumbersome. A more precise and compact way of describing the address space
was desired when assigning
CIDR blocks of addresses. The creation of the classless IP network address
system provided the network community with a new shorthand for writing IP
network masks.
In this shorthand, a forward slash, /, followed by the number of bits
set to one in the network mask, is used instead of the four-octet dotted-decimal
mask. A network mask of 255.255.0.0 has 16 bits of ones, so it can be written
as /16 (pronounced “slash
sixteen”). A network mask of 255.255.252.0 has 22 bits of ones, so
it can be written as /22. This type of mask is known
as a bitcount
mask. Combined with an IP network address, the network
shorthand of 131.108.0.0/16 can be used to represent 131.108.0.0 mask 255.255.0.0.
Likewise, 206.220.224.0/22 can be used to represent 206.220.224.0 mask 255.255.252.0
(which itself is a CIDR block representing the Class C addresses 206.220.224.0
through 206.220.227.0 each with
mask 255.255.255.0).
NOTE
During the System Configuration Dialog described in Chapter
2, “The Basics of Device Configuration,” it is assumed
that all network addresses fall along the classful network boundaries
described earlier. The Number of bits in subnet field
[0]: question that is posed to the user is asking how many bits
of the host component should be used for subnetting based on the class of
the network number the user entered. If the network number is a Class A network,
such as 17.0.0.0, 24 bits of host field could be used for subnetting. If the
user indicates that 9 bits are used for subnetting, the IOS calculates the
appropriate network mask, in this case 255.255.128.0.
Residential
Broadband, Second Edition){kind=link}
){kind=link}
){kind=link}