Configuring and optimizing multiple Exchange 2000 locations

Configuring Exchange 2000 on a stand-alone LAN is tricky enough, but when you start supporting multiple locations, things get interesting. In this Daily Feature, Walter Glenn shows you what's involved in supporting Exchange 2000 at multiple locations.

It doesn’t matter whether your Exchange 2000 organization spans countries, cities, buildings, or just different departments. The chances are that you will have to manage multiple locations. Exchange 2000 Server provides more flexibility in managing servers in disparate locations than ever, but this flexibility comes at the price of added complexity.

In previous versions of Exchange, a single construct, called the site, was used to group servers for routing and administrative purposes. Exchange 2000 uses two separate components, administrative groups and routing groups, instead of the site.

Exchange 2000 Server provides a flexible way to manage servers in different locations. At the physical level, you can group well-connected servers into routing groups and then connect those routing groups using different types of connectors. At the administrative level, you can group servers and other Exchange resources into administrative groups and then assign different permissions for administrators of those groups.

Administrative groups
Administrative groups are logical collections of Exchange objects that are grouped to facilitate administration and permissions management. Administrative groups can include servers, routing groups, public folder trees, system policies, and more. These administrative groups can be designed to fit your needs by designating geographical boundaries, departmental divisions, different groups of Exchange administrators, or different Exchange functions. For example, separate Exchange administrator groups might be responsible for managing the organization’s messaging backbone, public folders, and storage. Each administrative group would contain only the objects needed by those administrators.

Microsoft defines three basic administrative models for Exchange 2000: centralized, decentralized, and mixed.
  • Centralized: In the centralized model, one administrator or group of administrators maintains complete control over an entire Exchange organization. This does not necessarily mean that only one administrative group is defined, however. You might choose to create several groups to make it easier to assign permissions. This type of model is typically found in medium to large companies that have a centralized administrative staff.
  • Decentralized: In the decentralized model, each location (whether geographical, departmental, etc.) would have its own administrators and administrative group. This model is perfect for companies that maintain a single organization that spans large geographical distances and has separate administrative staff members in each location.
  • Mixed: The mixed model is simply a combination of the centralized and decentralized models and is probably a better example of what you find in the real world. A company with a basically centralized structure, for example, might open a branch location or acquire another company.

Administrative groups address a significant problem with earlier versions of Exchange: The administrative needs and the physical routing needs on a network do not necessarily match.

Routing groups
Routing groups are physical groupings of Exchange servers that have full-time, full-mesh, reliable connections between each server in the group. Messages sent between any two servers within a routing group are delivered directly to the destination server from the source server.

In an ideal world, all servers on a network would be well connected. But this is not usually the case. For this reason, Exchange 2000 lets you define routing groups full of well-connected servers and then connect those routing groups to each another using several different kinds of connectors.

Routing groups in Exchange 2000 are planned according to available bandwidth and reliability of the connection, a strategy similar to sites in earlier versions of Exchange. However, because Exchange 2000 uses SMTP instead of X.400 as the default transport protocol, Exchange 2000 is more tolerant than previous versions of lower bandwidths and higher latency. This means you’ll be able to group servers into the same routing group, a strategy that may not have been possible with an Exchange site.

The most important factor to consider when defining routing group boundaries is the stability of the network connection rather than the actual bandwidth of the connection. If the connection is prone to failure or is often too saturated with network traffic to be useful, then you should divide your servers into separate routing groups.

Connecting routing groups
Once you have created multiple routing groups, you will need to connect them so that they can exchange information. There are three types of connectors that you can create: the Routing Group Connector (RGC), the SMTP connector, and the X.400 connector.
  • Routing group connector: The RGC is the primary connector used to connect routing groups and is the fastest and easiest to configure. It is also reliable. It uses SMTP as its default transport mechanism, but it may also use a Remote Procedure Call (RPC) if the situation requires it, such as when connecting to a server running a previous version of Exchange. The RGC can be configured to use specific bridgehead servers (servers designated to be the transmitter of messages over the connection—see Figure A) or to use all servers in the routing group as bridgehead servers.

  • Figure A
    Bridgehead servers connect different Exchange 2000 locations.

    • SMTP connector: The SMTP connector takes a bit more work to set up than the RGC and sports different features. It is mainly used to connect routing groups where you want to force SMTP to be used for the transport mechanism. The SMTP connector is also easier to configure than the RGC and offers better options for fine-tuning the connection. The SMTP can also be configured to use multiple bridgehead servers.
    • X.400 connector: The X.400 connector can also be used to connect routing groups. X.400 connectors are useful for linking routing groups when there is very little bandwidth (less than 16 KB) available between servers, when the connection is unreliable, or when X.400 is the only connectivity available. When linking routing groups with the X.400 connector, a single server in each group must be designated as the bridgehead server. You must set up multiple X.400 connectors between multiple servers in each routing group to gain a load-balancing feature.

    Your best choice, in most cases, is the RGC because it is the easiest to configure and manage. However, your choice will depend on the purpose the connector will serve. In addition, multiple connectors may be created between the same two routing groups to provide a level of fault tolerance and load balancing.

    When creating multiple connectors, it is often useful to assign different cost values to the connectors to bias the use of certain connectors over others. The cost value indicates the cost of using the connector relative to any other connectors that may connect two routing groups. This value can range from one to 100, and lower cost links are always preferred over higher cost links. For example, you might want to configure two connectors to share the main messaging load between two routing groups and assign both of them a cost of one. You might then configure a backup connector with a cost of five that is used when the two main connectors are unavailable (see Figure B).

    Figure B
    You can have multiple connectors and assign different costs to each.

    Editor's Picks