Networking → Does IP subnet size/broadcast domain affect efficiency?

Should be easy question for networking gurus--
I have 100 devices on LAN, 172.16.0.0/16
If I change the IP scheme to 172.16.0.0/24, ALL else being equal, does that have any effect on network efficiency?
Thanks

No.

Won't make any difference as long as the number of devices is the same.

Don't forget that 172.16.0.255 is your broadcast address for that network with that subnet mask). As for efficiency, I think 254 would be good for a home network router. Any bigger and you might need the big guns.

All you're doing is going from one subnet mask to another. So...you're going from 255.255.0.0 to 255.255.255.0. This will give you a network 172.16.0.0 with 254 usable hosts (this includes the router). Usable ranges are 172.16.0.1-172.16.0.254 (if you have 172.16.0.1 assigned to the router then you really only have 253 addresses left for your network. You're basically sub-netting your class B address with a class C subnet mask.

Thanks for all the responses - I suppose I could explain my "need to know." I inherited a flat network using 172.16.0.0/16 with about 500 hosts and a lot of static IPs.
I'm preparing to separate into two or more VLANs, and I intended to move each host into new VLANs with about 100 hosts going into 172.10.10.0/24 and the rest into 172.20.20.0/22. Then, I decided to keep it simple to begin with, I'd keep the first part at 172.16.0.0/16, and now I am wondering if there's any need to move it at all. I thought I'd read that a smaller broadcast domain is more efficient, but I now think it was referring to the total # of actual hosts, not the size of the IP space (potential hosts), which is what my brain understood.
Thanks for confirming.

Ooops. I sort of had the idea right. You wouldn't be subnetting at all with 255.255.0.0 since that's a default mask for a B class network. So in that case your host range for that network would be 172.16.0.1-172.16.255.254 Your broadcast address would be 172.16.255.255 in that case.

The more hosts, generally the more broadcast traffic on the network. It's not really a problem until you can barely do anything if anything at all on the network. It might slow down the overall throughput of the network during heavy usage but the only way to find out is to try it. That /16 block will give you a B class subnet mask allowing your network increments to go up by 16. So in your case:

172.16.0.0
172.16.16.0
172.16.32.0
172.16.48.0
and so on...

Now you are changing things. You said 100 hosts and "everything being the same".

What becomes an issue if you split the network is do all hosts need to talk to one another. If, so then you add the overhead of a router between subnets.

A smaller broadcast domain will help some. I have seen larger numbers of machines. The broadcasts were still not a large percentage, but make debugging connections more difficult as you have a lot of useless chatter when tracing network traffic.

If the hosts on the two VLANs do not need to talk to each other, then 2 VLANs with the same /16 subnet and the router on both VLANs will work. It makes it simpler if you have some common machines like servers; they can be members of both VLANs.

(All my enterprise networking is a decade out of date)

Large subnets were especially bad when we still used hubs, every packet was seen by every device and only one device at a time could transmit. If you add to this PCs using NetBIOS, which is unroutable, you get a lot of contention, not much data transfer. Netware and AppleTalk were 'chatty' protocols, that transmitted packets no one was waiting for.

Today, no device should be plugged into a hub in a network, ideally it is all switches. This makes almost all unicast packets go to a single port, no user device not interested ever sees it. (If you use one large switch rather than several small switches, it is possible no infrastructure link sees unicasts either)

TCP/IP won the networking wars, almost no one is intentionally using other protocols. (I would check your network printers sometime, many still have chatty protocols turned on by default, to ease installation, I bet you are not deliberately using AppleTalk, VINES or Netware today)

Pure TCP/IP is not very chatty, although if you use Microsoft File and Printer sharing over it it becomes more chatty. Any SNMP Network management program you have should be able to tell you what percentage of your traffic is broadcasts, and what percentage is non TCP/IP.

If you bring both of those to near zero, there is no downside to a 'flat' network. (I first used VLANs because both my firm's Novell servers had the same name, they became very confused when we built a site wide network)

Sorry for the confusion - I explained some of my thought process in the second post, probably unnecessarily. The real question is whether it is wise to change the existing subnet mask from /16 to /24 when I remove all but 100 hosts, or if it really makes no difference. My confusion was whether the size of a broadcast domain was measured by the number of possible hosts (almost 65k with /16) or by the number of actual hosts.

I'm no expert in subnetting, but I don't think that's right. 172.16.0.0/16 gives you IPs from 172.16.0.1 to 172.16.255.254. What you've described is 172.16.0.0/20. I do my subnet calculations at this site: »www.t1shopper.com/tools/ ··· -subnet/

thx

Does that include shared printers from a server OS, such as Windows 2003 server?I'm creating the VLANs to separate school admin traffic and servers from student access, while still sharing the Internet connection. There needs to be no traffic across the VLANs. I have had a low-priority VLAN for guest wireless access for a while now, and that has worked well. Thx

Yes, you can 'browse' for Microsoft printers because they are chatty, they announce their presence every so often if any one cares or not.

To print to a TCP/IP printer, you have to know it's IP address already, it never advertises.

depending on how you are segmenting the network into vlans (i.e. using something like a cisco switching running a layer3 image), if you don't put any access-rules into place, then the devices will just "route" between the two vlans.
most of the time that vlan segmentation is implemented is to either (a) shrink the broadcast domain or (b) segment the network for restricted access to one or more portions of the network. normally, (a) is the case, especially if you are running ip phones or cheap switches (with small cam tables for mac-address storage), however, if you are *really* interested in preventing access to certain parts of the network, make sure that your router/layer3 switch includes the correct access rules to prevent unauthorised access.

q.

No layer 3 - the main router is a Peplink, but it is only for the three WAN connections, feeds into a PIX serving as firewall and VPN endpoint. The network is mostly aged HP Procurve with a Procurve 2824 serving as 'core' switch. It has some limited routing capabilities, can have a few static routes, but by default does no routing between VLANs, which is just fine for this application. BTW, the VLANs are port-based.
I do have a related question, which I should probably post anew, but I'll give this a shot. Everything is now on the default VLAN (ID: 1). After I move all the student/faculty to their own VLAN/subnet (i.e. VLAN ID: 20), is it a good idea to move the 172.16.0.0/16 school admin subnet to its own VLAN (i.e. VLAN ID: 10) and not use the default VLAN? I've read contradictory opinions about whether there is any security benefit to doing that.
TIA