Subnet Howto

Here's a complete list of all the subnet sizes you can have in IPv4, from smallest to largest:
You'll also need an IP for the default router (if it's just a stub network), so the number of IP's usable for hosts is total - 3.

To simplify things in the following examples it's only going to use last 8 bits of the network address and netmask rather than the full 32.

N.B. you don't have to read this bit if maths scares you, use the table above.

But first I'd better tell you how to convert from binary to decimal and back.

This is actually pretty simple once you get the hang if it. The first thing to do is to cast your mind back to when you where learning basic arithmetic and where learning about the 'ones, tens, hundreds, and thousands column'

In base 10 (decimal) each column goes up in powers of 10:

In base 2 (binary) each column goes up in powers of 2:

Converting from base 2:

Lets say we've got 00110011 and we want to convert it to decimal

Write down all the 8 powers of two from 128 to one from right to left, and write the bits underneath:

Now, take all the numbers in the top row that are over a 1, and add them together:

32 + 16 + 2 + 1 = 51

So 00110011 in decimal is 51

Converting back again is pretty easy too, just take the number and divide it by the highest power of two and record the answer. Then take the remainder and divide it by the next highest and so on, either until the remainder is zero or you get to 1:

If you look down the results column (after the equals) you get 00110011, Which is what we started with...

Now, your probably wondering how this is relevant to networking and how the network address and netmask interact, and how you use the host and network portions of the netmask.

Well it's

The next thing you need to know is an operation called a binary AND, it's also pretty simple
 

Because the netmask must always be a continues line of '1' bits from right to left, each subsequent 1 bit produces two networks that are half the size of the previous one and fit exactly inside it.

This is hard to visualize

• Firstly the number of IP addresses in a subnet will always be a power of 2 (2, 4, 8, 16, 32, 64, 128, etc...)

• And secondly the boundaries that subnets can occur on are fixed.

The second point could probably do with some explanation. probably the best way is to show you this table which shows you all the ways you can split a /24 (a Class C). 
 

You can see that if you used the first 64 IP's for one subnet (a /26), and the last for a second, you couldn't then use the middle 128 IP's for one subnet, you'd have to make 2 more 64 IP subnets or move both /26's to the beginning of the range or at the end.

e.g.

Here are some easy things to check to make sure that you've got the subnet split properly:

• The network address will always be an even number.

• The default router/gateway, which by convention is the network address + 1 will always be an odd number.

• The broadcast address is always an odd number.

• The highest usable IP address (broadcast address - 1) will always be an even number.

Note that in any IP network (apart from unnumbered point to point links) you lose 2 IP's - one for the network address (the smallest one), and one for the broadcast address (the highest one). You can use the network address for a host under some circumstances, but it's best to leave it alone.

If you want to find out more about this then type 'CIDR' into your favorite search engine, there is plenty of info out there.