The idea behind clustering is to have two or more servers working together to keep an application up and running. The application is actually hosted in a "virtual server" -- with its own name, IP address and shared storage. The underlying structure is transparent to users -- all they see is the application as it would be hosted on a single server.
The one feature lacking today with Exchange Server clustering, as shipped in Windows Server 2003 and Exchange Server 2003, is the fact that at any given time the shared data -- the information store -- is not being replicated, and thus becomes a single point of failure.
You can gain that functionality by using third-party products like NSI's DoubleTake that replicates data on a byte level, and additionally provides support for failing over to remote locations. The next release of Microsoft Exchange is reported to include store replication as well.
Do you have comments on this Ask the Expert Q&A? Let us know.
This was first published in November 2005