Planning for DHCP implementation
Adequate planning prevents headaches with DHCP installation, so find out how best to implement DHCP to avoid performance and security problems on your network.
If you're planning on upgrading to Windows 2000, Microsoft recommends using TCP/IP as the default protocol and implementing TCP/IP through Dynamic Host Configuration Protocol (DHCP). However, before you install DHCP on your network, you must consider some serious configuration issues. In this article, I'll discuss some of these issues. As I do, I'll provide you with information on how best to implement DHCP to avoid performance and security problems on your network.
If you've been using TCP/IP for any time at all, you're no doubt familiar with DHCP. If not, here's a brief explanation: DHCP is a TCP/IP service that you can use to set network clients' TCP/IP configurations automatically as they attach to the network. This configuration is done by providing the DHCP server with one or more scopes. A scope is nothing more than a range of IP addresses that may be assigned to clients on a temporary basis. The length of time that a client is allowed to use an IP address from a DHCP server is called the client's lease period.
As you can see, DHCP eliminates the need to manually configure each client; it also helps to prevent duplicate IP addresses and can help to conserve IP addresses when they're in short supply. However, for DHCP to function correctly and efficiently on large networks, proper planning is essential. If it is set up improperly, DHCP can cause headaches. The following sections discuss some of the issues you need to consider before installing DHCP.
As I mentioned earlier, DHCP automatically configures clients for TCP/IP at the time the clients attach to the network. Obviously, this process doesn't happen by magic. Getting DHCP to configure a client for use on a network requires several messages to be passed back and forth across the network. Although these messages tend to be small, if many clients are attempting to get an IP address at the same time, the excessive traffic created can bog down the network.
One way of dealing with this problem is to limit each DHCP server to servicing only one specific subnet. By doing so, you'll isolate each subnet's DHCP-related traffic and prevent it from flowing onto the rest of the network.
Of course, everyone is on a budget these days, and DHCP servers cost money. If you lack the funding to implement all those individual DHCP servers, it's possible to make a single DHCP server service multiple subnets. However, you must first consider some issues beyond network traffic. The biggest issue involves your routers. DHCP servers function through the use of broadcasts; therefore, before a DHCP server can service multiple subnets, the routers that connect those subnets must be configured as BOOTP and DHCP relay agents.
If your routers don't support BOOTP and DHCP, you're not out of luck just yet. If you have at least one Windows 2000 server or Windows NT 4.0 Server in each subnet, you can configure those servers to act as DHCP and BOOTP relay agents. Should you decide to use relay agents, keep in mind that it's often necessary to increase the default amount of time that relay agents wait before sending relay messages to servers.
As you can see, setting up a DHCP server to service multiple subnets can be a pain to configure. Doing so also has a negative impact on network performance. I recommend always limiting a DHCP server to servicing its own subnet unless absolutely necessary.