Take time to assess your current environment and to document the topology-not only Exchange but also your Active Directory® environment. You'll probably be pleased to know that Exchange 2007 has done away with link state routing and the topology based on routing groups and connectors. Instead, Exchange 2007 takes advantage of the investment your organization has already made in designing the Active Directory topology. As a result, all Exchange recipient and configuration data is stored in Active Directory, the routing topology is derived from the Active Directory site configuration, and all server roles will use site awareness to discover the services running on other server roles. Make sure that the current Active Directory site topology truly leverages the underlying physical network and that all Active Directory sites have associated subnets, then document the existing Active Directory sites and IP site links. Also document where all of your Exchange 2003 servers are physically located, and the routing group and routing group connector structure. This is the trickiest part of the transition process: moving from an Exchange routing infrastructure based on routing groups to a routing infrastructure based on Active Directory sites. For a small organization, it's generally a simple process. But a large organization with many routing groups will need to plan carefully for an interim phase when Exchange 2003 and Exchange 2007 are coexisting. The last thing you want is to end up having a message sent to your Exchange 2003 mailbox from the Exchange 2007 user across the hall routed through some remote routing group over a low bandwidth connection. Let's start by making some assumptions about your current topology: The Exchange 2003 organization is running in native mode. You have more than one routing group. You have more than one Active Directory site.
Figure:diagrams an Exchange and Active Directory topology.
How Exchange 2007 Uses Active Directory
When an Exchange 2007 server starts, it is stamped with a site attribute that helps other Exchange 2007 servers locate the services provided by that server. Only the Hub Transport server can use SMTP to transport a message within the organization. Each Active Directory site that contains a Mailbox server must also contain a Hub Transport server and, if the mailbox users access their mailbox by using any non-MAPI method, each site must also contain a Client Access server. Any time that a message needs to be processed for delivery, it will pass through a Hub Transport server, which will make a decision about how the message should be routed. If the message is destined for a Mailbox server in the same Active Directory site as the Hub Transport server, the Hub Transport server will deliver the message to the mailbox. If the message is destined for a Mailbox server that's in a different site, the Hub Transport server will relay the message directly to a Hub Transport server in the remote site.
The Hub Transport server uses the Active Directory IP site link cost information to calculate the lowest-cost route to the Active Directory site where the recipient mailbox is located. The route selection algorithm is very similar to Exchange 2003, but it's based on IP site link costs instead of routing group connector costs. Moreover, the message does not stop at each Hub Transport server along the way. It goes directly from source to destination. So why does it bother to calculate the lowest-cost route if it's relying on the IP network to transport the message? There are a couple of reasons. One is to delay message bifurcation. A message that is being sent to more than one recipient may need to be delivered to Mailbox servers in more than one Active Directory site. Rather than bifurcate, or split, the message at the first Hub Transport server, Exchange 2007 will not split the message until it reaches a fork in the routing path. As a result, the message will be relayed directly to a Hub Transport server in the Active Directory site that represents the bifurcation point. This behavior is known as delayed fan-out.
The lowest cost route is also used to determine where to queue the message in case the destination can't be reached. If a Hub Transport server in the target Active Directory site can't be reached, the sending Hub Transport server will then attempt delivery to a Hub Transport server in the next closest Active Directory site according to the routing path. Message delivery will continue along the lowest cost route until it reaches an Active Directory site where a Hub Transport server is available. Finally, if no Hub Transport servers along the route to the recipient Active Directory site are available, the message is queued locally. This method queues the message as close to the delivery point as possible, helping to make diagnosis of network failures more deterministic. This behavior is known as queue-at-point-of-failure.
Exchange 2003 works in a completely different manner. It calculates the lowest-cost route from one routing group to another based on the costs assigned to the routing group connectors. A bridgehead server in each routing group along the routing path will receive and then relay the message. If the next connector in the path is not available, an attempt is made to calculate an alternative route. Link state update messages are also communicated throughout the Exchange organization to notify the other Exchange servers that the connection is down. The bridgehead servers will attempt to route around the down connector until a link state notification is received indicating that the connection is up.
The challenge when transitioning a large organization is to maintain mail flow during the coexistence period. To achieve this continuity when Exchange 2007 is introduced into the environment, all Exchange 2007 servers become members of a single routing group. This means that regardless of which Active Directory site the Exchange 2007 server is in, Exchange 2003 will see it as belonging to that single routing group. This allows you to establish a routing group connector between that routing group and the Exchange 2003 routing groups so that Exchange 2003 can figure out how to route messages to Exchange 2007. Exchange 2007 will also use the routing group connector to determine how to get messages to Exchange 2003. However, Exchange 2007 will always prefer to route a message through another Exchange 2007 server, and will never backbone across an Exchange 2003 routing group to reach another Exchange 2007 server.
When an Exchange 2007 server starts, it is stamped with a site attribute that helps other Exchange 2007 servers locate the services provided by that server. Only the Hub Transport server can use SMTP to transport a message within the organization. Each Active Directory site that contains a Mailbox server must also contain a Hub Transport server and, if the mailbox users access their mailbox by using any non-MAPI method, each site must also contain a Client Access server. Any time that a message needs to be processed for delivery, it will pass through a Hub Transport server, which will make a decision about how the message should be routed. If the message is destined for a Mailbox server in the same Active Directory site as the Hub Transport server, the Hub Transport server will deliver the message to the mailbox. If the message is destined for a Mailbox server that's in a different site, the Hub Transport server will relay the message directly to a Hub Transport server in the remote site.
The Hub Transport server uses the Active Directory IP site link cost information to calculate the lowest-cost route to the Active Directory site where the recipient mailbox is located. The route selection algorithm is very similar to Exchange 2003, but it's based on IP site link costs instead of routing group connector costs. Moreover, the message does not stop at each Hub Transport server along the way. It goes directly from source to destination. So why does it bother to calculate the lowest-cost route if it's relying on the IP network to transport the message? There are a couple of reasons. One is to delay message bifurcation. A message that is being sent to more than one recipient may need to be delivered to Mailbox servers in more than one Active Directory site. Rather than bifurcate, or split, the message at the first Hub Transport server, Exchange 2007 will not split the message until it reaches a fork in the routing path. As a result, the message will be relayed directly to a Hub Transport server in the Active Directory site that represents the bifurcation point. This behavior is known as delayed fan-out.
The lowest cost route is also used to determine where to queue the message in case the destination can't be reached. If a Hub Transport server in the target Active Directory site can't be reached, the sending Hub Transport server will then attempt delivery to a Hub Transport server in the next closest Active Directory site according to the routing path. Message delivery will continue along the lowest cost route until it reaches an Active Directory site where a Hub Transport server is available. Finally, if no Hub Transport servers along the route to the recipient Active Directory site are available, the message is queued locally. This method queues the message as close to the delivery point as possible, helping to make diagnosis of network failures more deterministic. This behavior is known as queue-at-point-of-failure.
Exchange 2003 works in a completely different manner. It calculates the lowest-cost route from one routing group to another based on the costs assigned to the routing group connectors. A bridgehead server in each routing group along the routing path will receive and then relay the message. If the next connector in the path is not available, an attempt is made to calculate an alternative route. Link state update messages are also communicated throughout the Exchange organization to notify the other Exchange servers that the connection is down. The bridgehead servers will attempt to route around the down connector until a link state notification is received indicating that the connection is up.
The challenge when transitioning a large organization is to maintain mail flow during the coexistence period. To achieve this continuity when Exchange 2007 is introduced into the environment, all Exchange 2007 servers become members of a single routing group. This means that regardless of which Active Directory site the Exchange 2007 server is in, Exchange 2003 will see it as belonging to that single routing group. This allows you to establish a routing group connector between that routing group and the Exchange 2003 routing groups so that Exchange 2003 can figure out how to route messages to Exchange 2007. Exchange 2007 will also use the routing group connector to determine how to get messages to Exchange 2003. However, Exchange 2007 will always prefer to route a message through another Exchange 2007 server, and will never backbone across an Exchange 2003 routing group to reach another Exchange 2007 server.
No comments:
Post a Comment