This alert is server generated alert, means the DB triggers the alert and EM just receives and displays it.
If there is no "GC wait" event reporting in AWR Top 5 wait events or If you are not facing any performance issue. These alert can be ignore safely....
You can disable/change the thresholds value of this Metric to reduce the frequency of this OEM alert under below navigation for each instance of RAC
1) From EM console => Navigate to CLFDB_CLFDB2 DB home page => oracle database => monitoring => Metrics and collection settings => locate "Global Cache Statistics"
2) Change the value
2) Save the changes.
It can be beneficial to look at V$SYSSTAT for the statistic "global cache cr blocks received". If you have a low number of blocks being transferred across the cluster interconnect, then a high value for the average get time can be indicative of a slow private network. (it is private, right?).
If you have a high value for the blocks received, then your application may be relying on to much cross-instance Cache Fusion traffic. Application partitioning can help this if you have multiple applications