Why is there a difference between ONTAP "statistics volume" and Active IQ Unified Manager IOP counts?
Applies to
- OnCommand Unified Manager ( OCUM )
- Active IQ Unified Manager ( AIQUM )
- ONTAP 9
Answer
- The volume layer, or WAFL layer, is a maximum of 64kB in size.
- The protocol (LUN or CIFS/NFS operations) can be bigger (65kB to megabytes in size) in operation size.
- Multiple operations are broken up into parallel 64kB requests between the protocol and WAFL.
- This may also impact a QoS policy:
- QoS policies engage at protocol layer operation counts.
- Example: Microsoft SQL may do database queries on a LUN in 256kB sequential reads which translate to 4 64kB WAFL read messages.
- Another reason is because the workload may come in on different nodes, so the IOP count will depend on the node being measured against if using the CLI output of
qos statistics
Additional Information
- To compare the two, you can use the following commands:
Cluster::> set diag -c off; statistics start -object volume-vserver <SVM name> -instance <volume name> -node <node where data LIF is>
Cluster::> qos statistics volume performance show -vserver <SVM name> -volume <volume name>
- NetApp Harvest volume view shows metrics from the WAFL layer (similar to
statistics start -object volume
).
Example: Screenshots of IOPs from Harvest and AIQUM