Real-World Replication Designs

Site topology in Windows Server 2003's Active Directory has been engineered in a way to be adaptable to network environments of all shapes and sizes. Because so many WAN topologies exist, a subsequently large number of site topologies can be designed to match the WAN environment. Despite the variations, several common site topologies are implemented, roughly following the two design models detailed in the following sections. These real-world models detail how the Windows Server 2003 AD site topology can be used effectively.

Hub-and-Spoke Replication Design

CompanyA is a glass manufacturer with a central factory and headquarters located in Leuven, Belgium. Four smaller manufacturing facilities are located in Marseille, Brussels, Amsterdam, and Krakow. WAN traffic follows a typical hub-and-spoke pattern, as diagrammed in Figure 7.19.

Figure 7.19. CompanyA WAN diagram.

CompanyA decided to deploy Windows Server 2003 to all its branch locations and allocated several domain controllers for each location. Sites in Active Directory were designated for each major location within the company and given names to match their physical location. Site links were created to correspond with the WAN link locations, and their replication schedules were closely tied with WAN utilization levels on the links themselves. The result was a Windows Server 2003 Active Directory site diagram that looks similar to Figure 7.20.

Figure 7.20. CompanyA site topology.

Both domain controllers in each site were designated as a preferred bridgehead server to lessen the replication load on the global catalog servers in the remote sites. However, the PDC Emulator in the main site was left off the list of preferred bridgehead servers to lessen the load on that server. Site link bridging was kept activated because there was no specific need to turn off this functionality.

This design left CompanyA with a relatively simple but robust replication model that it could easily modify at a future time as WAN infrastructure changes.

Decentralized Replication Design

CompanyB is a mining and mineral extraction corporation that has central locations in Duluth, Charleston, and Cheyenne. Several branch locations are distributed across the continental United States. Its WAN diagram utilizes multiple WAN links, with various connection speeds, as diagrammed in Figure 7.21.

Figure 7.21. CompanyB WAN diagram.

CompanyB recently implemented Windows Server 2003 Active Directory across its infrastructure. The three main locations consist of five Active Directory domain controllers and two global catalog servers. The smaller sites utilize one or two domain controllers for each site, depending on the size. Each server setup in the remote sites was installed using the Install from Media option because the WAN links were not robust enough to handle the site traffic that a full DCPromo operation would involve.

A site link design scheme, like the one shown in Figure 7.22, was chosen to take into account the multiple routes that the WAN topology provides. This design scheme provides for a degree of redundancy, as well, because replication traffic could continue to succeed even if one of the major WAN links was down.

Figure 7.22. CompanyB site topology.

Each smaller site was designated to cache universal group membership because bandwidth was at a minimum and CompanyB wanted to reduce replication traffic to the lowest levels possible, while keeping user logins and directory access prompt. In addition, traffic on the site links to the smaller sites was scheduled to occur only at hour intervals in the evening so that it did not interfere with regular WAN traffic during business hours.

Each domain controller in the smaller sites was designated as a preferred bridgehead server. In the larger sites, three domain controllers with extra processor capacity were designated as the preferred bridgehead servers for their respective sites to offload the extra processing load from the other domain controllers in those sites.

This design left CompanyB with a robust method of throttling replication traffic to its slower WAN links, but at the same time maintaining a distributed directory service environment that AD provides.