Exchange 2000 determines the route a message takes based on a
least cost algorithm. Each Exchange 2000 server has a map of the
entire messaging topology of which it is a member. This map,
represented in the
Exchange uses a link propagation protocol called the
After a link is marked as down, the original routing continues to retry the connection at 60-second intervals. Even though no message is waiting to transfer, the routing keeps trying to contact the destination server. Once a connection has been reestablished, the routing notifies the local routing group master that the connection is available and the routing group master notifies all servers in the routing group that the connection is available.
Link state information is most effective when multiple routing groups are configured in an organization, particularly if redundant paths are available. Each routing group has a master that is fed link state information from different sources. The master keeps track of the link state data and propagates it to the rest of the servers within the routing group. The master is normally the first server installed in the routing group, but you can change this in System Manager by going to the routing group, clicking Members, right-clicking the server, and then selecting Set as Master. When a non-master receives new link state information, it immediately transfers the link state information to the master, so that other servers can receive the information about the routing change.
Only two states exist for any given link, up or down, so connection information such as whether a link is active or in a retry state is not propagated and is confined to the server involved in the message transfer.
When a server in a routing group determines that the state of a connector assigned to it has changed (gone up or down), it attempts to update the master's link state information with the new information. If the master is unavailable, it cannot update this information. The result is that servers in the organization continue to operate with old (possibly incorrect) information about connectors in the routing group whose master is unavailable. Information about connectors in other routing groups continues to be updated as the connectors go up and down, and even servers in the routing group with the failed master continue to receive information about the state of other routing group as mail is transferred.
The network operates less efficiently when the master is down, and every reasonable attempt should be made to make the master available to all servers in the routing group at all times. It is important to note, however, that even if the routing group master is down, all servers continue to operate with loop-free link state data.