When using Real application Clusters (RAC), Each instance has its own SGA and buffer cache. RAC will ensure that these block changes are co-ordinated to maximize performance and to ensure data intergrity. Each copy of the buffer also called as a cache resource has a master which is one of the nodes of the cluster.
In database releases before 10g (10.1.0.2) once a cache resource is mastered on an instance, a re-mastering or a change in the master would take place only during a reconfiguration that would happen automatically during both normal operations like instance startup or instance shutdown or abnormal events like Node eviction by Cluster Manager. So if Node B is the master of a cache resource, this resource will remain mastered on Node B until reconfiguration.
10g introduces a concept of resource remastering via DRM. With DRM a resource can be re-mastered on another node say from Node B to Node A if it is found that the cache resource is accessed more frequently from Node A. A reconfiguration is no longer the only reason for a resource to be re-mastered.
In 10gR1 DRM is driven by affinity of files and in 10gR2 it is based on objects.
DRM attributes are intentionally undocumented since they may change depending on the version. These attributes should not be changed without discussing with Support.
Two instance will not start a DRM operation at the same time however lmd,lms,lmon processes from all instances collectively take part in the DRM operation.
Normal activity on the database is not affected due to DRM. This means users continue insert/update/delete operations without any interruptions. Also DRM operations complete very quickly.
DRM many cause" latch: cache buffers chains" and "latch: object queue header operation " wait event, you can go throught this way to disable DRM:
_gc_affinity_time=0
_gc_undo_affinity=FALSE
also, you can used another two implicit parameters dynamic change
_gc_affinity_limit=250
_gc_affinity_minimum=10485760
You can set this parameter larger .