I am currently running a single GPU Cluster Compute node on Amazon EC2 which is listed as having 22 GB of RAM per instance.
top confirms that the instance does indeed have 22 GB of RAM but on an idle systems it appears that 5-6 GB of RAM are consistently in use. Neither
ps aux can identify the processes to which this usage is attributable.
5-6 GB seems too large for the OS to me, but perhaps I am wrong. Could this be an overhead due to virtualization on Amazon? Any other ideas or suggestions as to how I could track this down?