Subnet Size Recommendations

James Fry
James Fry used Ask the Experts™
on
All, quick question.  For a long time, I have always seen that a class C subnet should be .24 or smaller.  However, I've never really seen a reason or reasons why that is.  Any insight from anyone?
Comment
Watch Question

Do more with

Expert Office
EXPERT OFFICE® is a registered trademark of EXPERTS EXCHANGE®
Classes are not really used so much any more as what is called CIDR. The smaller the subnet the last traffic less broadcast traffic, etc.

Now a /22 or /21 won't kill you, but there will be more traffic. It will depend upon the devices and how chatty they are as to how many you'd want on the same subnet.  If they all blab alot then you'd get more congested.
Darrell PorterEnterprise Business Process Architect
Commented:
Class A addresses have their first octet in the range 1 to 126 (binary address begins with 0).

Class B addresses have their first octet in the range 128 to 191 (binary address begins with 10).

Class C addresses have their first octet in the range 192 to 223 (binary address begins with 110).

Sub-allocation of a Class C, using subnet masks of 25, 26, 27, 28, 29, and 30 bit subnet masks is used when it is known a network will never grow beyond the limits imposed by the number of IP addresses.

For example, when a router is connecting to another router and to no other devices then a /30 (2 usable IP addresses) is used to ensure a simple WAN link doesn't absorb an entire Class C subnet.

Your ISP may give you a block of 5 "usable" IP addresses, with the 6th IP address being the default router as assigned by the ISP.  Your ISP would issue you a block with a /29 subnet mask.

Some networks with a large number of client workstations and printers may need more than 253 usable IP addresses so may opt to use a /23 bit subnet mask to allow for 509 useable IP addresses.

To determine the optimal configuration of the subnet mask, the primary factor is to determine how large the network will grow.  If you believe a network will never need more than 253 routable IPs than use a /24 bit subnet mask.  If you want to maintain simplicity in the IP schemas, standardize on the /24 bit mask.  If you have a number of point-to-point WAN links, use a single class C network and sub-allocate it using /30 bit subnet masks.
Don JohnstonInstructor
Top Expert 2015

Commented:
For a long time, I have always seen that a class C subnet should be .24 or smaller.  However, I've never really seen a reason or reasons why that is.
Since a class C address has an 8-bit host field, and if you subnet it, the host field would be smaller, that would be why.
Well, that's /24 or "smaller" subnets as in /25, /26,/27, etc.  Class C subnets can't be bigger - as in /23.

But, the use of "classes" is only an irritating anachronism for me.  The subnet masks don't care - just 1's and 0's that are in contiguous groups.
If you are using a /24 say for 172.16.8.0/24, just don't use 9.0/24, 10.0/24 or 11.0/24, and if you ever need to increase them you just change to /23 or /22/ or even /21 and gain more hosts in the subnet. I'm not sure I would go below /21 because that would be a ridiculous amount of traffic on one subnet, and I'd think about segregating and routing in that case or using VLANs.  The idea is: reserve the range of IPs after your subnet to be able to expand to it with a simple subnet mask change on all machines at a later date.  Also if you prefer, start at 8 or another boundary say 12 so you reserve the lower IPs for other things and also so that when you switch from /24 to /22 or /21 example, the reserved ranges will be the right ones.
You can't go from 8.0/24 to 8.0/21 and expect to use 6.0, 7.0,8.0 and 9.0 because a 8.0/21 would give you 8,9,10 and 11 so plan ahead and reserve appropriately for whatever beyond your wildest dreams might happen and you should be good.

A CIDR calculator is your friend: http://www.subnet-calculator.com/cidr.php

Of course you could always start with a /22 or 21 if you preferred and just not use all of the addresses if you don't need them.  I can't see where that would hurt a thing.  It isn't the amount of possible addresses as much as the amount of in use addresses that will generate broadcasts and such that could cause congestion. A good managed switch can do rate limiting to avoid a lot of the negatives of repeated chatter.

Do more with

Expert Office
Submit tech questions to Ask the Experts™ at any time to receive solutions, advice, and new ideas from leading industry professionals.

Start 7-Day Free Trial