Subnets and Subnet Masking

IP network addresses are given in the form <network><host> (or, if subnetted, <network><subnet><host>) when viewed in binary notation. Masks determine how much of the address is network, and how much host.

For a standard Class B address, the mask is 255.255.0.0, hence the first 2 octets are network and the last two are available for hosts (PCs, servers, etc.).

For Class C, the mask is 255.255.255.0, which means that only the last octet (256 addresses) is available for hosts.

In binary; the masks are:

Class B: 11111111.11111111.00000000.00000000

Class C: 11111111.11111111.11111111.00000000

Host numbers which are all zeroes or all ones are not allowed, hence a Class C address cannot be xxx.xxx.xxx.0 or xxx.xxx.xxx.255, giving 254 useable addresses.

Applying a mask.

A subnet mask increases the amount of address that is deemed to be "network" thereby reducing the amount of host for each address.

e.g. Class C address 194.64.48.86, mask 255.255.255.240

In binary:

Address 11000010.01000000.00110000.01010110

Mask 11111111.11111111.11111111.11110000

So the network is the 28 most significant bits.

To determine the subnet, the address is ANDed with the mask.

Address 11000010.01000000.00110000.01010110

Mask 11111111.11111111.11111111.11110000

=========================================

ANDed 11000010.01000000.00110000.01010000

Thus, to determine whether two addresses are in the same subnet, the address is ANDed with the mask, and the two results compared.

E.g. Is 194.64.48.86 in the same subnet as 194.64.48.162, when a mask of 255.255.255.240 is applied?

Address 11000010.01000000.00110000.01010110 194.64.48.86

Mask 11111111.11111111.11111111.11110000 255.255.255.240

=======================================================

ANDed 11000010.01000000.00110000.01010000 194.64.48.80

Address 11000010.01000000.00110000.10100010 194.64.48.162

Mask 11111111.11111111.11111111.11110000 255.255.255.240

========================================================

ANDed 11000010.01000000.00110000.10100000 194.64.48.160

Subnets are 194.64.48.80 and 194.64.48.160, so they are not in the same subnet.

Only the last octet need really be considered when testing this - 195.200.21.50 is clearly not the same network!

Also, if the subnet is applied in this way, simply determining the least significant bit of the mask (in our case, 10000₂ or 16) gives the number of addresses in each subnet, so the addresses will be 0-15, 16-31, 32-47, 48-63, etc. As "all zero" and "all one" addresses are not allowed, 194.64.48.0, 194.64.48.15, 194.64.48.16, 194.64.48.31, 194.64.48.32, etc. are not legitimate when masked to 255.255.255.240 (in binary, these make the last 4 bits all zeroes or all ones; these are the host addresses reserved for network or broadcast).

Silly Subnets

It is possible (and legal!) to apply odd subnet masks to addresses, such as to subnet a Class B by 255.255.55.57 (In binary, 11111111.11111111.00110111.00111001) but this causes severe headaches as you need a scientific calculator to work out each and every address for the subnet - try ANDing this sort of mask!