Single digit I/O wait is likely not an issue.
AIX is scheduling the process because it has nothing else to do.
The production system I support sees that and it is also on solid state storage.
Ideas:
You might look at Monitoring disk I/O
OS file locking from writes can block reads in databases - The mount option CIO lets the the database deal with locking. This requires unmounting the filesystem and should not be done to filesystem containing executables.
Adding noatime as a mount option to a filesystem can be good and can be done without unmounting the filesystem.
Open a perfpmr with IBM.
------------------------------
Alexander Pettitt
------------------------------
Original Message:
Sent: Wed September 17, 2025 02:43 AM
From: MAILVAGANAM VELAYUTHAM
Subject: LPAR Performance Issue During Heavy Job Run - AIX 7.3
Hi All,
I have an LPAR that is ready for the new cutover, but its performance is sluggish during heavy job runs.
I'm observing a consistent iowait of around 4.5–5.8%, while the nmon report shows that CPU resources are not being fully utilized. Initially, I suspected a storage issue, but no latency was observed (FlashSystem 9500 with NVMe interface, end-to-end).
Any suggestions on why the CPU is not fully utilized during the job run would be greatly appreciated.
Thanks,
Shanker
------------------------------
MAILVAGANAM VELAYUTHAM
Storage Consultant -IBM
Crossroad
------------------------------