We have a process of using RMAN to restore a DB to a VM server by reading a backup image stored on a NFS share, then after the successful restoration, back the DB up to the same NFS share.
During the restoration, both latency and IOPS on this NFS share were very high, 25 ms/op and 1400 IOPS. During the backup, the latency was only about 1-2 ms/op, and about the same IOPS. Throughputs are about the same in both case as well.
I thought the RMAN read and write both is sequential I/O, I cannot explain why the read would cause such high latency.
Can somebody please explain to me?
Thank you!