Subnet divider: split a block into subnets

The formula, with a worked example

• By number of subnets N: borrow b bits, the smallest b with 2^b ≥ N. New prefix = old prefix + b.
• By hosts per subnet H: the new prefix is the longest one with 2^(32 − prefix) − 2 ≥ H (a /31 counts 2 and a /32 counts 1).
• Subnets made = 2^(new prefix − old prefix). Each one holds 2^(32 − new prefix) addresses.
• Subnet k starts at the block's network address + k × 2^(32 − new prefix).
• Worked example: 192.168.0.0/24 into 4 subnets borrows 2 bits (2^2 = 4), so the new prefix is /26: 4 subnets of 64 addresses, 62 usable hosts each, starting at .0, .64, .128 and .192.

Dividing by number of subnets

Splitting a block means moving the prefix to the right. Every bit you move it, which is called borrowing a bit from the host part, doubles the number of subnets and halves the size of each one. To get N subnets you borrow the smallest number of bits b that gives at least N, that is 2^b ≥ N.

For 4 subnets that is 2 bits, since 2^2 = 4. For 5 subnets 2 bits is not enough (only 4), so it takes 3 bits and you get 8 subnets. Subnets always come in powers of two when every one is the same size, so asking for 5 gives 8, and the extra 3 are spare. The divider says when it has rounded up like this.

Take 192.168.0.0/24 into 5. Borrowing 3 bits makes it a /27, so there are 8 subnets of 32 addresses each, 30 usable hosts in each. They start at .0, .32, .64, .96, .128, .160, .192 and .224.

Dividing by hosts per subnet

When the size of each subnet matters more than the count, divide by hosts instead. The divider finds the smallest subnet that still holds the number of hosts you enter, which is the longest prefix where 2^(32 − prefix) − 2 is at least that number, and then fits as many of those into the block as it can.

For 50 hosts in 192.168.10.0/24, a /27 gives 30 usable hosts, which is too few, and a /26 gives 62, which is enough. So the block splits into 4 subnets of /26. The second of them runs from 192.168.10.64 to 192.168.10.127, with usable hosts .65 to .126.

For 500 hosts in 10.0.0.0/16, a /24 gives 254 (too few) and a /23 gives 510, so the block makes 128 subnets of /23. The second one is 10.0.2.0/23, with a broadcast address of 10.0.3.255.

Reading the list of subnets

Each row is one subnet: its network address in CIDR form, its first and last usable host and its broadcast address. The subnets are back to back, so each network address is the previous broadcast address plus one. The list shows up to the first 64 subnets; the count and size above it always cover the whole split, and any subnet further down can be found by adding the block size.

To find subnet number k (counting from 0), multiply k by the subnet size and add it to the block's network address. With /26 subnets of 64 addresses, subnet 3 of 192.168.0.0/24 starts at 192.168.0.0 + 3 × 64 = 192.168.0.192.

Edge cases and limits

If you enter an address that is not the start of the block, such as 192.168.0.77/24, the divider uses the block it belongs to, 192.168.0.0/24. A block cannot be split past /32, so asking for more subnets than the block has addresses is an error, as is asking for more hosts than the whole block can hold.

Dividing down to /31 gives subnets of 2 addresses that count 2 usable each under RFC 3021, and dividing down to /32 gives single addresses. Asking for 2 hosts per subnet picks /31 for that reason, and asking for 1 picks /32.

This tool makes equal-sized subnets. Variable length subnet masking (VLSM) mixes sizes inside one block; you can do that by dividing the block once, then dividing one of the resulting subnets again here.