How to measure NFS read/write latency per second

Description

Can I monitor Network File System (NFS) read/write latency in seconds from the filer? Perfstat gathers protocol and NFS stats but these provide an average from the sample period. They do not show possible latency peaks.

Procedure

The stats command that is built into Data ONTAP 7G can be used to gather more granular data points. If -i is specified,the time interval per second will be displayed.

The command listed below will display the nfsv3 read and write latency from the filer. Note that this information is reported in microseconds.
Therefore, 1000us = 1ms.

filer> stats show -i 1 nfsv3
Instance nfsv3_ops nfsv3_read_l nfsv3_read_o nfsv3_write_ nfsv3_write_

                /s           us                      /s           us               /s
     nfs      5938      2160.09         2989        79.97         2949
     nfs      6500      1839.16         3240        84.88         3260
     nfs      7193      1466.20         3528        71.96         3665
     nfs      7731      1351.35         3842        79.43         3886
     nfs      8549      1059.71         4232        81.54         4317
     nfs      6630      1056.74         3296       417.07         3334
     nfs      5838      1180.69         2946       977.56         2892

nfsv3_ops = NFSv3 READ/WRITE ops measure per iteration
nfsv3_read_l = NFSv3 READ latency
nfsv3_read_o =  NFSv3 READ ops
nfsv3_write_ = NFSv3 write latency
nfsv3_write_ = NFSv3 write ops

Note: Due to a formatting limit (nfsv3_write_), the last character that should be in nfsv3_write_o and nfsv3_write_l is missing.