I have setup 11gR1 (22.214.171.124) database on 11gR2 clusterware with 3 node cluster and have started noticing the swapping issue although AMM is set to 164G.
Searching through oracle doc recommend using linux huge pages. This is a Datawarehouse database. Has any one experience huge pages.
AMM under Linux uses POSIX shared memory that relies on /dev/shm. By default, /dev/shm is limited to 50 % of physcial RAM and uses 4 KB memory pages.
How much physical RAM does the server have installed and how large is your Oracle SGA? However, using POSIX shared memroy on servers with more than 16 GB of RAM installed is not recommended. If your Oracle SGA is using that much memory you are wasting a lot of memory just for managing the amount of memory pages and cannot take advantage of the TLB cache to improve virtual address translation.
I suggest you look into setting up Kernel Hugepages, which use 2 MB pages under Linux and configure Oracle SGA using ASMM. You should also see a drastic increase in Oracle server performance. http://docs.oracle.com/cd/B28359_01/server.111/b32009/appi_vlm.htm.