That depends on what you are trying to accomplish.

Clustering is essentially a solution for high
availability, which means a hardware solution that keeps
your SQL Server functional.

Replication is typically used to distribute data and can
be used for fault tolerance.

The choice between clustering, replication, and other
solutions is how valuable your data is to you or how
expensive downtime is. The more valuable your data or the
more expensive downtime, the more likely you are to use a
clustering solution.

Replication also is limited in that it can't replicate
system object, but can replicate tables, procs, views,
indexed views, and UDF's. You can also massage your data
as it goes down the wire. However there is a latency
associated with it. The latency associated with failing
over clustering is a function of hardware and how long it
takes to apply the transaction log from one server to
another, which is not finite - in other words can be in
the order of minutes.



[quoted text, click to view]