We are having a similar issue in our evaluation of Veeam ONE. Veeam ONE is reporting a Cluster Memory Overcommitment alarm, but our VMs are using a total of only 696.12 GB (45.15%) of memory as reported by the Memory Performance Chart and graph, when our cluster has 1152GB installed.
We have a four-host Hyper-V cluster (Windows Server 2012), supporting 80 VMs, with cluster reserve set to 1 (meaning the cluster will support all VMs if one host is offline). Each host has 384GB of memory, so the cluster's supported total memory is (4 hosts x 384GB) - 384GB = 1152GB.
Our two largest VMs are using 32GB each, with the next 10 largest at 16GB, and the next 12 at 12GB. The other 56 range from 8GB down to 2GB each.
Additionally, we use Microsoft's System Center Virtual Machine Manager 2012 (SCVMM). VMM has reported no cluster overcommitment. And, if we add our SCVMM server to the Veeam ONE Monitor, the same cluster shows no overcommitment alarm. The alarm is only seen when Veeam ONE looks at the cluster directly.
It's my understanding that SCVMM 2012 has greatly improved algorithms over previous versions for determining overcommitment status. I'm hoping that Veeam ONE is not using algorithms similar to SCVMM 2008 which often falsely reported overcommitment.
We like the Veeam ONE product, but this could be a deal breaker in our evaluation if we can't fully trust the alarm capabilities.