Non-Octet-Aligned Subnets
So far the subnet boundary always fell inside the fourth octet. Real networks use prefixes like /13, /19, /22 whose boundaries fall in the second or third octet. The good news: it is the same block method, just applied to a different octet.
Boundaries do not have to align
Everything you learned in Lesson 3 still holds. The subnet address is the IP AND the mask; the broadcast is the next block boundary minus one. The only new wrinkle is that the magic octet — the octet where the boundary falls — may be the second or third octet instead of the fourth. Once you can name that octet, the arithmetic is identical.
A /13 borrows five bits beyond an octet; an /11 borrows three. Those prefixes allocate many thousands of addresses because their host bits span more than one octet. You will meet them when an ISP hands a customer a large block, or when a campus numbers a whole building.
Students often assume a larger prefix means less work. The opposite is true here: writing 10.1.2.99/13 out as thirty-two bits and ANDing every column is slow and error-prone. The block method trims the task to a single octet, because only one octet ever holds the split. Everything to the left of it is fixed, and everything to the right is filled with 0s for the network or 255s for the broadcast. That economy of effort is exactly why the block method is the exam standard.
Finding the split octet
The split octet is the octet containing the last network bit of the mask — equivalently, the first octet (from the left) whose mask value is not 255, except when the boundary lands exactly on an octet.
256 − (mask value in the split octet). Step 3. Round the IP’s split octet down to the nearest multiple of the increment — that is the network octet. Step 4. Broadcast = network rounded up to the next boundary, minus 1 (all following octets become 255).Two examples to fix the idea: for /22 the mask is 255.255.252.0; the split octet is the third (mask value 252), so the increment is 256 − 252 = 4. For /17 the mask is 255.255.128.0; the split octet is again the third (mask value 128), so the increment is 256 − 128 = 128. The earlier “increment 4 for /22” and “increment 128 for /17” fall straight out of the same rule.
The nested magic-number map
There is only one set of magic numbers: 128, 192, 224, 240, 248, 252, 254, 255, whose increments are 128, 64, 32, 16, 8, 4, 2, 1. The same value appears in different octets to produce different prefixes. “Nest” the map onto whichever octet the boundary falls in.
Worked: /13, /22, /19
Each example is worked with the block method and then confirmed with the count rule.
Example 1 — 10.1.2.99/13
Mask /13 = 255.248.0.0
Split octet = 2nd (mask 248), increment = 256 − 248 = 8
Multiples of 8 in octet 2: 0, 8, 16, ...
Octet 2 of the IP is 1 -> rounds down to 0
Network 10.0.0.0
Broadcast 10.7.255.255 (next boundary 8 -> 10.8.0.0, minus 1)
First 10.0.0.1
Last 10.7.255.254
Check: 10.0.0.0 to 10.7.255.255 spans 8 /8-sized chunks
= 8 x 65536 = 524288 addresses = 2^19 OK
Example 2 — 172.16.14.7/22
Mask /22 = 255.255.252.0
Split octet = 3rd (mask 252), increment = 256 − 252 = 4
Multiples of 4 in octet 3: 0,4,8,12,16,...
Octet 3 of the IP is 14 -> rounds down to 12
Network 172.16.12.0
Broadcast 172.16.15.255 (next boundary 16 -> 172.16.16.0, minus 1)
First 172.16.12.1
Last 172.16.15.254
Binary check of octet 3:
14 = 00001110
252 = 11111100
AND = 00001100 = 12 OK
Example 3 — 192.168.50.203/19
Mask /19 = 255.255.224.0
Split octet = 3rd (mask 224), increment = 256 − 224 = 32
Multiples of 32 in octet 3: 0,32,64,96,...
Octet 3 of the IP is 50 -> rounds down to 32
Network 192.168.32.0
Broadcast 192.168.63.255 (next boundary 64 -> 192.168.64.0, minus 1)
First 192.168.32.1
Last 192.168.63.254
Check: octet 3 spans 32..63 = 32 values x 256 = 8192 = 2^13
and h = 32 − 19 = 13 OK
Why 10.7.255.255? A /13 has 32 − 13 = 19 host bits. The network is 10.0.0.0; setting all 19 host bits to 1 gives 10.7.255.255 (octet 2 can run 0–7, then 255.255). The address 10.1.2.99 sits inside it because octet 2 = 1 is between 0 and 7.
The /22 covers four values in octet 3 (12, 13, 14, 15), which is 4 × 256 = 1024 addresses, matching 210 host addresses for h = 32 − 22 = 10.
.0 for the network and end at .255 for the broadcast. Only the split octet carries the interesting number.More drills
Run the block method and check each with the count rule.
Confirm the last row by hand: 172.16.200.10/21. Mask 255.255.248.0, increment 8. Octet 3 = 200, which is already a multiple of 8 (25 × 8 = 200), so the network is 172.16.200.0. The block covers 200–207, giving broadcast 172.16.207.255 and usable .200.1–.207.254 = 2046 addresses (211 − 2, since h = 11).
Key terms & sources
Non-octet-aligned
A subnet boundary that does not fall exactly between two octets.
Split octet
The octet containing the last network bit; where the mask is neither all 1s nor all 0s (or the boundary octet).
Increment
256 − mask value in the split octet; the spacing between subnet boundaries.
Nested magic map
The 128/192/224/240/248/252/254/255 values mapped to prefixes in whichever octet they appear.
- Fuller, V. & Li, T. RFC 4632: Classless Inter-Domain Routing (CIDR) (2006) (alignment of prefixes to bits, not octets). rfc-editor.org/rfc/rfc4632.
- Pummill, R. & Manning, B. RFC 1878: Variable Length Subnet Table For IPv4 (1995). rfc-editor.org/rfc/rfc1878.
- Cisco. CCNA (200-301) Exam Topics, IPv4 Addressing and Subnetting. cisco.com.
Review
Check what you have learned with these review questions.
Target exams: CCNA 200-301 v1.1 · CyberOps Associate 200-201. Last content verification: 2026-10-11.