You are correct in that SQL clustering is a fail-over solution, not a
scalability solution.

You can measure throughput on your SQL server a number of ways, one of the
best being to take a "black box" profile trace for an hour and play it back
at maximum speed.  The ratio of recording time vs. playback time wil tell
you how much capability you have left.  The downside is that this requires
taking your server offline for an extended period of time.  This is best
done on a new server before going live.

There are publicly available tools that can individually stress components
so you can find your maximum IO rate (http://www.iometer.org/).

In general, SQL servers scale up, not out.  That means purchasing a larger
box.  Some parts of your system may be offloaded onto a read-only system,
using replication or log shipping to synchronize them, but what can be moved
varies greatly by application.

Finally, you should purchase and install a good monitoring solution to track
performance metrics and alert when they reach certain thresholds.  Idera
makes a good basic one (www.idera.com) and Quest makes a very good complex
one (www.quest.com).  The SQLH2 toolkit from Microsoft (FREE!!) has
performance logging and basic reporting capability.  Measuring where you are
is the first step towards understanding where you need to be.

Geoff N. Hiten
Microsoft SQL Server MVP