Bandwidth usage and consumption is an ongoing issue in network environments. As network environments have become more complex and distributed, the need to control traffic over WAN links has become more and more important, and a typical concern with an Active Directory implementation is the way AD will handle replication between sites.
Replication is the process of sending update information between domain controllers. Because every domain controller in the Active Directory environment holds a copy of the Active Directory database, each domain controller must be updated when a change occurs. For example, if you change a user's password, that password change must be replicated to all other domain controllers to ensure that the database is accurate on all other domain controllers. Without replication, the Active Directory would quickly become a collection of useless, inaccurate data.
The Active Directory does a good job with replication data because changes made to the Active Directory are replicated on an attribute level. In the Active Directory, objects are made of attributes, or descriptors of that object. For example, a user account may have attributes such as user name, password, email address, phone number, and so on. When a change is made to an object, only the changed attribute must be replicated—not the entire object. For example, let's say you change a user's password. The only item that must be replicated to across the environment is the password attribute for that object—not the entire user account. So, the Active Directory is thrifty in terms of bandwidth usage because attribute replication converses as much bandwidth as possible.
The Active Directory provides two methods of replication within the Active Directory environment: intrasite replication and intersite replication. Intrasite replication is replication within an Active Directory site. The Active Directory uses sites to determine which IP subnets are considered "well-connected" in your environment. By your site definition, the Active Directory assumes that the IP subnets within a site are well connected and that bandwidth is considered freely available and inexpensive. Using this assumption, the Active Directory automatically configures connection objects between domain controllers within the site so that replication between domain controllers occurs frequently and without a schedule. As an administrator, there is nothing you must configure or manage with intrasite replication—the Active Directory takes care of it without your intervention.
Intersite replication, on the other hand, is a more difficult animal to configure. You help the Active Directory understand what WAN connections are available between your sites and how the Active Directory should manage data. The process of managing replication between sites is a lot like walking a balancing beam—you have to balance what you want in terms of replication with what your WAN links can physically manage. Without a doubt, the more replication that occurs between sites, the more accurate the database will be all of the time. In most environments, however, constant replication over WAN links is not a practical solution. Therefore, most administrators are faced with a trade-off between data accuracy and time. In Active Directory terms, the time required to replicate data from domain controller to domain controller and site to site is called latency. As an administrator, the trick is to find the best replication balance to manage traffic between sites while keeping database data as accurate as possible—in other words, you want to reduce latency as much as possible.
Intersite replication is based on site links. Sites are connected by some kind of WAN communication link. This may be something as grand as a T3 link, or as small as a VPN or modem connection. Depending on how your sites are connected, you configure site links in the Active Directory. These site links define the WAN connections that are between your sites. After the Active Director understands how your sites are linked together, you can then begin to implement control features that can help you find the balance between accurate data and latency that is right for your organization. Unfortunately, there are no hard and fast rules—the Active Directory is flexible enough to allow you to find the balance that is right for your needs. So, you need to understand the concepts and then spend some time experimenting to find the balance that is right for you.
Cost. Your initial management task is to assign a cost to each site link. The cost of each site link is based on an arbitrary number that you assign to it. The Active Directory uses this cost to determine which site links have precedent over other site links. Lower-cost site links are favored over higher-cost site links. For example, let's say I have a Boston site and a Houston site. I have a T1 WAN connection between the two sites, but I also have a backup VPN connection. For replication, I want to make certain that replication always occurs over the T1 link, unless the link is down. If the link is down, the VPN link can be used. So, in terms of cost, I might give the T1 site link a cost of 10, whereas the VPN link might have a cost of 50. Using these numbers, the Active Directory always uses the lowest-cost link over the higher-cost link. With the cost assignment, I can rest assured that my best WAN connection between the two sites will always be utilized first.
Schedule. Your next management tactic is to use schedules carefully and wisely. In intrasite replication, replication occurs frequently and without a schedule. In intersite replication, you can use a schedule to determine how often replication can occur. The idea is to provide a schedule that allows replication to occur as frequently as possible, but does not use too much bandwidth during peak network hours. For example, if you know that your environment uses a lot of intersite bandwidth between the work hours of 10:00 a.m. and 2:00 p.m., you might create a schedule that restricts replication during those hours. How you schedule replication is entirely up to you, but once again, you'll want to find that fine balance between data accuracy and latency that is acceptable for your network.
Servers. For best replication performance, you should have at least one global catalog server in each site. In reality, the global catalog server may increase replication traffic, but it decreases user traffic over the WAN link, which leaves more room for replication. Also, consider placing a DNS server in each site, and make sure that the site clients connect to that DNS server for service, which will help reduce DNS traffic over the WAN link.
As you might imagine, one of the best things you can do to manage site replication traffic is to sit down with a pencil and paper and carefully plan your Active Directory infrastructure. Careful planning and the application of site configuration knowledge on your part will help you develop sites and replication plans that will meet the needs of your network and give you the best replication speed possible.