1

We are currently primarily using a 10.1.1.0/24 subnet at our small company (~250 employees, ~90 computer users). We're seriously considering doing some VLANing / network segmentation and isolation, and we're very close to too many devices on our little 10.1.1.0/24 anyhow.

The question I have is, why wouldn't we just switch to using 10.X.0.0/16 subnets? This would give us plenty of different subnets (256 - no way we'll come up with that many different network segments to isolate) and way more than enough hosts per subnet (65534). But most of the examples I see online they tend to do "harder" things like 10.1.X.X/23 or 10.1.X.X/17. If 256 segments / subnets is more than you would ever need in the foreseeable future, is there some other reason to complicate matters and use a netmask other than the easy (because it is divisible by 8) /16?

subnet calculator

  • I'm not sure what you mean by harder. 10.0.0.0/16 is no easier or more difficult than 10.0.0.0/17 and 10.0.0.0/23. It's all calculated the exact same way. In any case, the rule of thumb is that your business needs to plan for 10x growth, and if your plan is comfortable and works with that, then your are good. The company I work for doesn't have enough private addressing, even using all three private address ranges, because of past sloppy choices, and it is much more difficult to go back and make changes than it is to do it correctly in the first place. – Ron Maupin Jun 5 '18 at 17:31
  • 1
    If your proposed plan accounts for your estimated growth then go for it. – joeqwerty Jun 5 '18 at 18:21
  • @RonMaupin - I guess I mean it is just easier to look at and grok / describe. You can just say "10.1.0.0/16 is our servers, 10.2.0.0/16 is our printers, etc., etc." and you don't even need a calculator to know where the next subnet starts and ends. – Adam Nofsinger Jun 5 '18 at 19:37
2

The main reason to not use big subnet is for preventing broastcast problem/traffic if too much host is alive.

Cisco recommend fewer than 500 hosts to prevent broadcast problem.

Per the CCNA's guide:

enter image description here

  • 1
    But this is only a problem if you actually have more than 500 hosts, correct? Not necessarily address space for more than 500 hosts. – Adam Nofsinger Jun 5 '18 at 19:41
  • 1
    Adam, Yag is rightly cautioning you about the potential of broadcast problems if you continue to add hosts to the broadcast domain, but a lot of people seem to misunderstand that. It's not the size of the network, but rather it's the number of hosts per broadcast domain. Naturally, a larger network allows for a larger broadcast domain... but if you have 500 hosts you have 500 hosts, whether your network accommodates 500 or 5,000. – joeqwerty Jun 5 '18 at 20:49
1

The Internet Protocol is primarily about routing to other networks. When you define your subnet you are stating when traffic is to be routed away through the default gateway and when it should be handled by the local network protocols.

When communication is to take place between two hosts within the same network segment, IP translates the addresses to the underlying (presumably) Ethernet protocol using ARP.

Some organisations or network providers use large “Ethernet” networks. You can do clever things using spanning tree protocol to create fault tolerance. You can implement network controls at the Ethernet level rather than IP level.

There are risks to having a large segment. A badly behaved host (or hosts) could generate broadcast traffic that saturates the segment. This would depend on what protocols are in use and how much broadcast traffic they generate. Depending on your switch, multicast traffic might not be handled efficiently either.

These are not necessarily show stoppers.

Historically some routing protocols could not distinguish between different subnets with a high level of granularity. Addresses were categorised as class “A”, class “B” or “C” which in turn defined the subnet mask length as /8, /16, or /24 respectively. This is more relevant if your addresses are public. This has since become less relevant as the availability of public IPs has decreased and firms have begun to adopt CIDR (classless inter-domain routing). Many IP registrars require most address assignees to have at least 50% utilisation which again encourages people to use smaller subnet sizes. Again, this is mainly relevant if you are using public address space.

One downside with some 10.x.x.x addressing is that if you have users connecting from remote networks, if they too use the same 10.x.x.x addresses, that can create some routing problems.

In conclusion - a large 10.x.x.x network is fine.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.