Subnets and CIDR Without the Headache

An IP address is two addresses in one

An IPv4 address looks like one number. It is really two, glued together. The first part names the network — the street, if you like. The last part names one host on that network — the house.

Here is the catch: nothing in 192.168.1.37 tells you where the street ends and the house begins. That split is a separate piece of information, and it is exactly what a subnet mask — or a CIDR prefix — supplies. Get the split wrong and two devices can sit side by side yet disagree about whether they are neighbours at all.

Once you see an address as network plus host, the rest of subnetting is arithmetic. Honestly, most of it is counting in powers of two.

CIDR is just counting bits

An IPv4 address is 32 bits. A CIDR prefix counts how many of those bits, from the left, belong to the network. /24 means 24 network bits and 8 host bits. /26 means 26 and 6. The slash number is the whole mask, compressed into one count.

The dotted mask says the same thing the long way. Write the prefix as 32 bits — that many ones, then zeros — group them in fours of eight, and convert each group. /24 becomes 255.255.255.0. /25 becomes 255.255.255.128. They are not two systems; one is shorthand for the other.

Spot the pattern. Each step up in the prefix halves the block. /24 holds 256 addresses, /25 holds 128, /26 holds 64. The formula is plain: total addresses = 2^{(32 − prefix)}. Usable hosts are normally two fewer — the reason is in the worked example below.

One rule keeps masks valid: the ones must be contiguous. 255.255.255.0 is fine. A mask with ones and zeros mixed up in the middle is not a mask at all, and a decent calculator will refuse it.

Worked example: 192.168.1.37/24

Take a real address you will meet in almost every home: 192.168.1.37/24. The /24 says the first three groups are the network and the last group is the host. So this device is host 37 on network 192.168.1.0.

Two addresses in every normal subnet are spoken for. The lowest, all host bits zero, is the network address — 192.168.1.0 here. It names the network itself and no device may take it. The highest, all host bits one, is the broadcast address — 192.168.1.255 — used to talk to every host at once. Neither can be assigned, which is why 256 addresses leave 254 usable hosts, running from 192.168.1.1 to 192.168.1.254.

The calculator below derives those the binary way — address AND mask gives the network, network OR the inverted mask gives broadcast — but you rarely need the binary. Hold the split in your head and the four numbers fall out: network, broadcast, first usable, last usable.

Splitting one network into smaller ones

Subnetting is just moving the divider right. Start with 192.168.1.0/24 and borrow one bit: you get two /25 networks, 192.168.1.0 and 192.168.1.128, each with 126 usable hosts. Borrow two bits and you get four /26 networks — .0, .64, .128 and .192 — each with 62 usable hosts.

Why bother? Separation, mostly. Put printers on one subnet, guests on another, and a misbehaving device disturbs only its own block. Smaller blocks also mean smaller broadcast chatter. The cost is waste: every subnet you carve pays the two-address tax, so carving a /24 into sixteen /28s burns 32 addresses on network and broadcast duty alone.

A quick sizing habit helps. Need about 50 hosts? A /26 gives 62 usable — enough, with a little headroom. Need 60? Still a /26, but tight. Need 70? You have outgrown it; step down to a /25. Always size for the hosts you will have next year, not the ones on today's spreadsheet.

Private ranges and two odd prefixes

Most addresses in this guide are private — set aside for use inside your own network and never routed on the public internet. There are three blocks: 10.0.0.0/8, 172.16.0.0/12 and 192.168.0.0/16. Your home router almost certainly hands out addresses from the last one. That is why your 192.168.1.37 can exist in millions of homes at once without a clash — each home is its own island, joined to the internet through NAT.

Two prefixes break the “minus two” rule. A /32 is a single address — one host, no network or broadcast waste — and route tables use it to point at exactly one machine. A /31 holds two addresses and, under RFC 3021, both are usable: on a point-to-point link between two routers there is nobody to broadcast to, so the old reservation buys nothing. The calculator below handles both cases instead of blindly subtracting two.

Read any subnet, step by step

1. Find the prefix. If you were given a dotted mask instead, count its one-bits — or read it off the table above.
2. Work out the block size: 2^{(32 − prefix)}. That is your total address count.
3. Take usable hosts as total minus two — unless the prefix is /31 or /32, where the edge rules above apply.
4. Set the host bits to zero for the network address, and to one for broadcast. Usable hosts live strictly between them.
5. Paste the address and prefix into the calculator below and compare every line — network, broadcast, range, wildcard mask. A mismatch usually means the split point, not the arithmetic, was off.

That is the whole headache, gone. A subnet is a block, a CIDR prefix is where the block is cut, and every other number is bookkeeping.