VLSM is basically #Subnetting the subnet. The process is almost identical to of Fix-Length Subnet Masking with the exception where it reduces \(BL\) to only include enough number of hosts in the network, that is \(2^{BL}\) should be close to \(\lg{n}\). The bits deducted from \(BL\), denoted as \(BD\) will be the bits that used to build up a subnet mask (though in reality it is counted as part of the prefix). \(2^{BD}\) will be the total number of subnets that could be subnetted from their parent.
Note: The minimum requirements for \(BL\) should be at least 2 bits since there are two addresses reversed for network address and broadcast address.
We could conserve even more IP addresses by subnetting the subnetted subnet if it is still too large and there are plenty of spaces left to include more hosts after one round of VLSM. It could build up a Hierarchical Network.
Not only is VLSM able to conserve IPv4 Address, it is also a more efficient way of using available address space.