I have a question about Veeam One and its take on VM sprawl/oversizing. Now i know there is discrepancy between guest OS (windows) resource utilization and actual VM utilization but this question relates to an SQL VM. We have a SQL server VM with 20GB and 4 CPU's ( was created before my time). Veeam One is reporting that this memory should be reduced to 8GB memory and 2 CPU's. Now this is quite a difference so i need some guidance really? How accurate is it ? Is there a way of Veeam One being told that this is running SQL server to adjust its findings maybe or is this pointless?
In a physical world this would be an ideal spec however trying to get across that you cannot apply these rules when it comes to virtualisation.
Also looking at vkernel but swaying towards Veeam One at the moment for simplicity.
The problem with SQL Server VMs is that these VMs will ALWAYS consume all amount of memory you've allocated to these VMs. To detect over-sized VMs we use Memory Usage counter, the memory which is actively used by the VM. Furthermore, in order not to be too aggressive with the recommendation we add addition 25% of buffer to the average memory usage you have for the VMs in question.
I would recommend taking a look at the performance graph of memory usage for this very VM (either via drill down report or a predefined VM performance report templae), I bet you will notice that it is not actively using all 20 GBs of RAM.
