Dynamic Host Configuration Protocol

Previous Topic Next Topic

Multihomed DHCP Servers

For a server computer to be multihomed, each network connection must attach the computer to more than one physical network. This requires that additional hardware (in the form of multiple installed network adapters) be used on the computer.

A computer running Windows 2000 Server can function as a multihomed DHCP server. The DHCP server binds to the first IP address configured on each network connection (that is, each physical adapter interface) in use on the server. By default, the service binding depends on whether the connection is dynamically or statically configured for TCP/IP. If statically, the connection is enabled in the binding to listen to and provide service to DHCP clients. If dynamically, it is disabled in service bindings and does not provide service to DHCP clients. Dynamic configuration methods include the use of either another DHCP server to obtain a leased IP configuration or self-configuring an address through the use of the APIPA feature provided in Windows 2000. For more information, see "DHCP and Automatic Private IP Addressing" earlier in this chapter.

Server scopes use the primary IP address for each multihomed network connection to communicate with the DHCP clients. To verify the primary IP address for each of the connections used in a multihomed server configuration, you can review the Internet Protocol (TCP/IP) properties for each connection listed in the Network and Dial-up Connections folder on the server.

Configuring a Multihomed DHCP Server

Figure 4.23 is an example of a multihomed DHCP server with three network adapters installed. Each adapter is configured to lease addresses on separate physical subnets.

Figure 4.23    Multihomed DHCP Server Configuration
Enlarge figure

Figure 4.23 Multihomed DHCP Server Configuration

The multihomed DHCP server has three adapters installed and configured statically with a single IP address for each. Because the IP addressing for the DHCP server also uses an adjusted or custom subnet mask value (255.255.255.224), that value is applied for all the IP addresses that are configured at the server and for other computers in use on the same network. Here, a Class C range of IP addresses, 192.168.200.1 to 192.168.200.254, is used.

Each of the three adapters connects the server to three different physical subnets (Subnets A, B and C). To achieve the intended results of having the DHCP server provide leased configuration service to all clients in each of the respective subnets, two configuration details are essential and must be verified during deployment plans:

  1. The server must use a statically configured IP address within the same range of valid IP addresses for the physical network on which it is servicing clients.
  2. The server must have each of its valid subnet IP addresses excluded from the scope used to offer leases to clients.

For example, if no special subnetting was used in this environment, the selection of DHCP server IP addresses is not as critical because the IP network and IP subnet are the same. When the default subnet mask value (255.255.255.255) for this example network is applied and in use, all 254 possible computer IDs are considered part of one single unified subnet.

If, however, a custom subnet mask of 255.255.255.224 is applied, the network ID and subnet ID are not the same. When the subnet ID is not the same as the network ID, make sure the DHCP server is provided an IP address assignment within the same subnet it is meant to service.

For instance, with the mask set to 255.255.255.224 at all computers, the first 3 bit places of the last notated octet (224) are taken from the full 8 bit places that would normally comprise the full computer ID section. These bits are used by IP for physical subnet identification. This leaves the remaining 8 bit places to be used as the actual or reduced computer ID field.

In this way, the example network shown above requires of the three subnets in use that they have a maximum of eight (or 23) potentially different subnet IDs. Likewise, each of these subnets can, in turn, only support up to 32 (or 25) potential computer IDs.

Because of the use of subnetting, Subnet A in this example consists of the first 32 address values in the network, from 0 to 31. Because no computer IDs consisting of all 0s or all 1s in the computer ID field can be assigned for use to computers, the useful range of total available IP addresses for each subnet drops from 32 to 30.

Of the remaining 30 addresses, the DHCP server needs to use one. The remaining 29 can be configured in a regular DHCP scope and used for assigning leases to subnet clients. The choice of which address to use for the DHCP server is at the administrator's preference, as well as the decision to either include the DHCP server's statically assigned IP address within the scope defined for use in each subnet.

The multihomed server's IP addresses (192.169.200.1, 192.168.200.33, 192.168.200.65) are configured using the first IP address available for use in each of the three subnets. For the configuration shown, these addresses are excluded from the defined boundaries of each of the scopes created for use with these subnets.

Alternatively, you can set up your scopes to include these addresses within the defined boundaries of the scope. If you do, you need to create address exclusions to exclude these server IP addresses from each of the respective scopes.

If more than a single IP address is statically configured for a network connection, the Windows 2000 DHCP Server service permits only the first configured IP address to be used in the context of enabling or disabling service bindings.

© 1985-2000 Microsoft Corporation. All rights reserved.