I'm seeing a (to my eyes) puzzling scenario where my 3090G2 NAS cluster seems to have a ratio of 2:1 r/w ops, so that for one write op I have two read ops. The same is true for ethernet bandwidth, but not for FC bandwidth (which is 1:1). See attached pictures. What you see is a storage migration done from my ESX5.1 environment (thin provisioned VM disks), moving from one datastore (EVS on the first NAS) to the second datastore (EVS on the second NAS). This is why one NAS seems to only read and the other to only write. (If I migrate the EVSs around, say to the same NAS, the ratio of 2:1 r/w ops is still seen).
Can anyone explain the asymmetric r/w ops scenario please?
More tech stuff: I have two 3090G2s metro clustered. Underlying storage is VSP. Each NAS node is connected to its own VSP - the VSPs then run True Copy on the underlying disks. So this means all the disks are in one I/O group, though stretched across two datacenters. The distance is negligible, around 100 meters. I'm running jumbo framed, and do NOT have the Jet Clone (VAAI) license for the NAS, if relevant.
I'm new to HNAS, so I apologize if my question qualifies as "doh, that's normal".