Design rules for Active Directory sites

Sites are an extremely useful design element for Active Directory domains. Sites are limited to any computer object within a forest. Thus, they can cross domains and organizational units (OUs) with indifference. An object's membership in a domain or OU does not exclude simultaneous membership in a site. Sites are used to impose physical network divisions for the purpose of traffic flow. The type of control sites offer includes restriction of authentication traffic and resource queries to DCs within the site, as well as controlling replication. Intra-site replication occurs automatically on demand between all site-member DCs (i.e., normally). However, inter-site replication is restricted to a defined time schedule and bandwidth throttle.

By definition sites are collections of IP subnets that have fast and reliable communication links between all hosts. Another way of putting this is a site contains LAN connections, but not WAN connections, with the general understanding that WAN connections are significantly slower and less reliable than LAN connections. By using sites, you can control and reduce the amount of traffic that flows over your slower WAN links. This can result in more efficient traffic flow for productivity tasks. It can also serve to keep WAN link costs down on the pay-by-the-bit services.

In general, when designing sites, keep the following in mind:

• Sites should generally reflect the physical or geographic topology of the network.
• Each site should contain at least one local DC.
• Sites should not contain slow links of any type.
• Remote-access clients do not need a dedicated site.
• Sites should be used whenever control over replication traffic is needed or desired.
• Sites can be added, removed, changed, and moved easily without affecting any other AD container configuration.

James Michael Stewart is a partner and researcher for ITinfopros, a technology-focused writing and training organization.