A Little History
In earlier versions of Exchange, multiple redundant servers could be used for particular roles, such as front-end or bridgehead servers, but there was no easy way to create redundant mailboxes or public folder servers. Windows Clustering Services could be used, but the actual mailbox and public folder data still resided in a single set of files accessed by whichever cluster node was active. You still had a single point of failure (the shared disk), and no protection against corrupt data, requiring third party solutions, such as Doubletake or NeverFail, to eliminate single points of failure. While these can be valid solutions, they are complex to implement and include the concern about supportability from Microsoft.

New Solutions
Exchange 2007 introduces two new High Availability features, with a 3rd on the way. These features all use a method called log shipping to keep independent replicas of Exchange 2007 storage groups (i.e. groups of mailbox files). These features are:

• Local Continuous Replication, which creates one or more replicas of storage groups on a single non-clustered mailbox server. Logs from the mailbox server are copied and shipped to a different location on the same server, preferably on separate physical disk partitions. A copy of the mail itself is then created in the second location by replaying the log file. Mechanisms to prevent corrupted data from being replayed to the replica are included. If the primary disks for the information store fail or the storage group becomes corrupt, an administrator can manually switch to the replica storage group.
• Cluster Continuous Replication, which provides even greater storage group availability by shipping the logs to an entirely different server. There is no single point of failure because you have two servers, each with its own set of disks. Cluster Continuous Replication requires the use of Windows Clustering Services (part of Windows 2003 Enterprise Edition) between the two servers. In conjunction with the log shipping you also get the advantages of clustering with automatic service monitoring and failover, and replication of Exchange configuration settings.
• Standby Continuous Replication is expected to be included with Service Pack 1 for Exchange 2007, which will hopefully be available by the end of the year. With Standby Continuous Replication data can be replicated on a per-storage-group basis to your choice of either standby servers or clusters. The targets, whether a single mailbox servers or a clusters, can be placed either inside the primary data center or in a remote location, ready to be manually activated if the primary server or data center fails.

Our Recommendations
Given the choices, what does Coyote Creek Consulting recommend?

If you need to roll out soon, use Cluster Continuous Replication. If you’re not in a rush, or if you don’t want to use Windows Clustering Services, wait and see if Standby Continuous Replication delivers on its promises. What about Local Continuous Replication? Because it does not include the failsafe of truly redundant servers, this service really doesn’t make sense for most users. We only recommend Local Continuous Replication for those who cannot afford to purchase a second server but still want to have a way to recover if the database gets corrupted.

High Availability means that your system can be up 99.99% of the time. If you have specific Service Level Agreements (SLA’s) in place which you are not meeting, Exchange 2007’s High Availability features can be a solution.

Back to News & Events