New high availability features in Exchange Server 2010

If you've used Exchange 2007's high availability features, you know that implementing HA can be painful. For example, if you wanted to cluster a mailbox server, you must create that cluster up front and it can only host the mailbox server role. If you start out with a non-clustered mailbox server and decided you wanted to use clustering, there is no way to cluster the server. Instead, you have to build a separate clustered mailbox server and move all of your mailboxes from the old server to the new clustered server.

Exchange 2010 improves on how the server handles high availability. You no longer have to cluster your mailbox servers at the time of deployment. In fact, Exchange 2010 completely does away with the concept of a clustered mailbox server.
A new feature -- incremental deployment -- lets you implement site resilience and high availability on an as needed basis. Furthermore, you can structure your Exchange 2010 organization so that failover (or switchover) occurs at the database level rather than at the server level, although server level failover and switchover is still an option.

Exchange Server 2007 allows you to host up to 50 databases on a single mailbox server. If a problem occurs with a database, you need to failover the entire server -- even though 49 of the databases were still healthy.
In Exchange 2010, failover and switchover occur at the database level. If a database failure occurs, you can activate the database on another server that contains a replica of the database -- without affecting any of the functional databases that are hosted on the server.

By redesigning the way that clustering works in Exchange 2010, Microsoft has also improved Exchange Server's scalability. Now, instead of being limited to hosting 50 databases per server, a single server can host up to 100 databases.
Exchange 2010 also improves upon cluster size allowance. Cluster continuous replication (CCR) limited Exchange 2007 mailbox servers to only two copies of a database -- one active and one passive. In contrast, Exchange 2010 clusters can have up to 16 copies of a database -- one active and up to 15 passive.

But this raises a few questions. Do you really need 16 copies of an Exchange database? Most organizations won't need to create a cluster that large. Now you can increase the size of your cluster as your business continuity requirements change.
Additionally, databases are no longer tied to a specific server. Exchange 2010 makes it simple to move the active copy of a database from one server to another. Therefore, you won't have servers that only act as hosts for passive database copies. In fact, Microsoft recommends that larger organizations scatter active database copies multiple mailbox servers rather than having one server that contains all active databases and other servers that contain only passive databases.

Since Exchange 2010 is only in beta, anything could change by the official release date. I don't expect Microsoft to make any major architectural changes to Exchange 2010, it's possible that clustering recommendations could change. In the meantime, I recommend that organizations that are thinking about adopting Exchange 2010 consider how this new architecture will affect Exchange Server backups.