Then you can have more networks and less hosts / network.
Our example IP address:
172.16.0.1
Subnet w/ 3 octets for the network:
255.255.255.0
A subnet / IP is 32 bits, right?
3 octets = 24 bits (8*3)
Ok, stay with me now. We're going to use the formula
(2^N-2) to figure out how many networks & hosts that will leave us with.
N = number of masked bits in used in the subnet. (In our case, this is 8 since the normal class B subnet uses 16.)
(2^8)-2 =254 subnets
Now we do the same for hosts / subnet:
(2^8)-2 = 254 hosts / subnet
In this case, N = the remaining number of bits in the subnet. So you have 8 bits left. Why? 32 bit - 16 bits (for normal cB subnet) = 16 bits - 8 bits (for subnet mask) = 8 bits.
Basically, you just further divide your network for organizational purposes.
-----------------------------------------------------------------
Another way to look at it:
IP: 172.16.0.1
Normally the "172.16" part would represent your network. The "0.1" part would represent your hosts. That means you could have
2^16-2=65,534 hosts on this single "172.16" network.
Instead, you've changed it to where "172.16.s" represent the network. So now only the ".1" represents your hosts. This means
2^8-2=254 host can be on this single "172.16.0" network... but, since you've changed the subnet by adding an extra octet to it, the "s" in "172.16.s" can be any number between 1-255. So that gives you another
2^8-2=254 networks.
Broken down...
Normal:
IP: 172.16.0.1
Subnet: 255.255.0.0
# of networks: 1
# of hosts / network: 65,634 (hopefully I did my math right)
Modified:
IP: 172.16.s.1
Subnet: 255.255.255.0
# of networks: 254
# of hosts / network: 254
Hopefully I didn't just confuse you more... lol.