We have a virtualized server on a NetApp SAN, SQL Server 2014, Windows Server 2012.
It is oversubscribed, in the sense it does not have enough memory.
When the low memory conditions occur and load is put on disk, the following counters rise above 25:
Avg Disk sec/Read
Avg Disk sec/Transfer
Of course, memory needs to be addressed, but I find it troubling that when the disks are put under load the transfers appear to be slow.
Our SAN administrator replies:
"The disk transfers mean nothing when the system is under load. We’re using software Iscsi which is highly susceptible to degradation due to lack of server resources. When the server is slow the disks can appear slow, but in reality the disks are performing well within normal tolerances. This is simply a symptom of virtualization and resource contention."
Does the SAN administrators response sound reasonable?