When tasked with IP address assignment for a customer's network, you need a plan. That plan needs a strategy to make the most efficient use of space, but it also needs flexibility to accommodate changing business needs. Let's look at some factors to consider when developing your plan.
Networks vs. hosts
Obviously, given the variable subnet size of TCP/IP, one of the most important things to consider is how large you should make the subnets. You need to consider how many subnets you need, and how many hosts go in each.
Note: Theories about the best way to do this have evolved as networking hardware has evolved. At one time, there was a general preference (humorously attributed to the "Flat Earth Society") for enormous subnets since routers were so slow and such bottlenecks. With the advent of layer-3 switching, the pendulum swung the other way to "micro-subnetting" which was supposed to minimize broadcast traffic. Since most of the protocols causing problems with broadcasts, like IPX, Appletalk, etc., are becoming rare these days, there isn't a burning need for lots of tiny subnets.
Internal vs. external
Almost every customer will need Internet access and have a DMZ type area. So, your plan needs to take into account addresses that will be visible from the Internet and addresses that will only be reachable internally.
Public vs. private
Many organizations own registered IP space, and even more get registered allocations from their providers. Where and how you choose to use this resource can affect a lot of things, such as whether you have to use NAT or whether you have to readdress everything if you change Internet providers.
VoIP vs. data
Many IPT manufacturers suggest separating VoIP and data traffic into different VLANs, which necessarily means different subnets. If you choose to take this approach, it should be included in your plan.
Also keep in mind it's popular to segment many other types of devices. These include subnets for server farms, and backup and out-of-band management networks.
One of the most important things you can do to make routing easy to manage is to keep the addressing easy to summarize. Ideally, all the subnets from a given office on the network should come from a single summary address.
Inside each subnet, it's also common to create conventions for which addresses are used for what devices. For example, gateways are usually .1 or .254 in a /24 subnet. You may also decide to have a portion of the subnet's addresses reserved for static assignment and the remainder assigned via DHCP.
The important part of creating a flexible plan is to consider these things even if the customer doesn't currently have a requirement for them. For instance, the customer may not have plans to deploy VoIP, but if the IP addressing scheme is set in stone and the customer decides to deploy VoIP next year, it will create problems.
About the author
Tom Lancaster, CCIE# 8829 CNX# 1105, is a consultant with 15 years of experience in the networking industry. He is co-author of several books on networking, most recently,CCSP: Secure PIX and Secure VPN Study Guide, published by Sybex.