Oracle internals, debugging and undocumented features
If you’re seeing high cluster waits because you have multiple processes all inserting into the same set of tables, the cluster waits are resolved by pointing to a single node on the cluster. But just pointing to a single node defeats the rollover benefits of running on RAC. To that end, Oracle gives us Services to assist with workload management. It’s a TNS entry that <em>prefers</em> a specific node, but will use a secondary instance if the preferred is unavailable. So all processes that use a common set of resources can all be pointed to the same instance, eliminating the cluster waits.
In practice I have seen this improve performance by 500% on high volume inserts, so definitely a feature <a href=”http://docs.oracle.com/cd/B19306_01/rac.102/b14197/hafeats.htm#CHDGEBED” title=”worth checking out.” target=”_blank”></a>