Since connections between sites have limited bandwidth, we prefer to schedule replication to occur at specific times when network activity is low. Domain controllers will be updated across our WAN links, but latency becomes a factor. Transmissions between sites can use RPC over IP for synchronous connections, or SMTP for asynchronous, unreliable links. SMTP can send schema, configuration, and Global Catalog updates, but it cannot send data for the directory partition, since the File Replication Service (FRS), which is responsible for global policy data, is limited to synchronous updates. To facilitate the process, data is compressed.
Wide area networks are not ordinarily configured point to point. A router is usually connected to a telecom vendor's central office for a particular service, such as frame relay. The layout pictured in Figure 14.4 indicates all sites connected with 512Kbps links. Using this diagram, you can see that each office can contact the others. This is referred to as a fully transitive system.
Site links represent the replication routes between two or more sites. Each link can be assigned a relative cost and a schedule. By default, the update frequency is three hours. Figure 14.4 illustrates the physical connections for three sites. For this example, you would configure one instance for the replication process, as you would schedule a conference call.
The KCC generates a spanning tree to supervise inter-site replication. The first server created at a site is assigned the role of Inter-Site Topology Generator (ISTG). Its function is to create a site topology to coordinate internal and external replication traffic. If that server becomes unavailable, the next oldest domain controller at that site is promoted to ISTG status.
Bridgehead servers are gateways used to connect adjacent sites. If you have more than one domain controller at a particular site, you might want to assign one to be a bridgehead server to act as a conduit, communicating with domain controllers on other sites and exchanging updates with partners within its site. This takes the load off the other servers.
Transitive site links are created automatically, so that all site links are bridged, as in Figure 14.4. But, if you need to control part of your network because one of your links is slower than the others, or it can only connect with one site, you can turn off automatic bridging and create a site link bridge. This is a common situation with branch offices and satellite locations where fast links must be balanced with slower ones.
Was this article helpful?