You are on page 1of 6

BRINGING IP NETWORKS TOGETHER

i. Now that we have some of the fundamentals of TCP/IP under our belt, let’s take a look
ii. at how they work to let us glue networks together.

iii. In this section, we cover the differences between hosts and networks, netmasks, static routing, and
some basics in dynamic routing.

iv. The purpose of this section is not to show you how to configure a Linux router, but
v. to introduce the concepts.

vi. Although you may find it less exciting than actually playing, you’ll find that understanding the basics
makes playing a little more interesting.

vii. More importantly, should you be looking to apply for a Linux system administrator’s
viii. job, these could be things that pop up as part of the interview questions.
Hosts and Networks
The Internet is a large group of interconnected networks.

All of these networks have agreed to connect with some other network, thus allowing
everyone to connect to one
another.

Each of these component networks is assigned a network address. Traditionally, in a 32-bit IP


address, the network component typically takes up 8, 16, or 24 bits to encode a class A, B, or
C network, respectively.

Since the remainder of the bits in the IP address is used to enumerate the host within the
network, the fewer bits
that are used to describe the network, the more bits are available to enumerate the hosts.

For example, class A networks have 24 bits left for the host component, which means
there can be upward of 16,777,214 hosts within that network. (Classes B and C have
65,534 and 254 nodes, respectively.)
The ranges are as follows:
Class Octet Range

A 0–126
B 128–192.167
C 192.169–223

You probably noted some gaps in the ranges. This is because there are some special
addresses that are reserved for special uses. The first special address is one you are likely
to be familiar with: 127.0.0.1. This is also known as the loopback address. It is set up on
every host using IP so that it can refer to itself. It seems a bit odd to do it this way, but just
because a system is capable of speaking IP doesn’t mean it has an IP address allocated to
it! On the other hand, the 127.0.0.1 address is virtually guaranteed. (If it isn’t there, more
likely than not, something has gone wrong.)
Three other ranges are notable: Every IP address in the 10.0.0.0 network, the 172.16–
172.31 networks, and the 192.168 network is considered a private IP. These ranges are not
allowed to be allocated to anyone on the Internet, and, therefore, you may use them on
your internal networks.
Subnetting
Imagine a network with a few thousand hosts on it, which is not unreasonable in a
medium-sized company. Trying to tie them all together into a single large network would
probably lead you to pull out all your hair, bang your head on the wall, or possibly both.
And that’s just the figurative stuff.

280 Linux Administration: A Beginner’s Guide

The reasons for not keeping a network as a single large entity range from technical
issues to political ones. On the technical front, there are limitations to every technology
on how large a network can get before it becomes too large. Ethernet, for instance, cannot
have more than 1024 hosts on a single collision domain. Realistically, having more than
a dozen on an even mildly busy network will cause serious performance issues. Even
migrating hosts to switches doesn’t solve the entire problem, since switches, too, have
limitations on how many hosts they can deal with.
Of course, you’re likely to run into management issues before you hit limitations of
switches; managing a single large network is difficult. Furthermore, as an organization
grows, individual departments will begin compartmentalizing. Human resources is usu-
ally the first candidate to need a secure network of its own so that nosy engineers don’t

peek into things they shouldn’t. In order to support a need like that, you need to create
subnetworks, a task more commonly referred to as subnetting.
Assuming our corporate network is 10.0.0.0, we could subnet it by setting up smaller
class C networks within it, such as 10.1.1.0, 10.1.2.0, 10.1.3.0, and so on. These smaller
networks would have 24-bit network components and 8-bit host components. Since the
first 8 bits would be used to identify our corporate network, we could use the remaining

16 bits of the network component to specify the subnet, giving us 65,534 possible subnet-
works. Of course, you don’t have to use all of them!

You might also like