>>>> I am not sure what equates to "response time". Is it "Average Service Time" or "Average Wait Time" or a combination of the two? <<<<
You have a good question. Disk response time normally means service time + wait time. If you see a lot of wait time in both msec. and frequency of its occurrence that also correspond to occurrence of high disk IOPS, you need more disk units. Your case above is not bad yet because the chart indicates that your disk IO workload of MB/sec. is generally very low. If disk IO workload increases in the future, and you see more wait time in conjunction with high IOPS, it's time to either add more disk units or if you have a lot of SQL/Query workload, to look at index advisor of large physical files and create useful indexes which will help reduce total disk IO workload and therefore disk response time.
That bad response time of about 9 msec. at 10:30 AM should not be a big concern as it lasted 15 minutes or less and I suspect it may happen due to a burst of high disk IOPS. You should produce the Physical Disk IO Overview - Basic chart as well to see disk IOPS profile and use it in association with Disk Throughput Overview for Disk Pools chart. Then look at them again after PowerHA is in operation or when users complain of slow workload performance.
Original Message:
Sent: Mon October 07, 2024 12:51 PM
From: John Paisley
Subject: Removing disk units from configuration using SST
Thanks again Satid. Based on the chart below, which was taken on a typical busy day, Wed, Oct 2nd, it looks like we are in pretty good shape for the time being - pre-PowerHA. Although, I am not sure what equates to "response time". Is it "Average Service Time" or "Average Wait Time" or a combination of the two?
Thanks,
John
------------------------------
John Paisley
Original Message:
Sent: Fri October 04, 2024 09:06 PM
From: Satid S
Subject: Removing disk units from configuration using SST
Dear John
>>>> You might want to run a performance analysis before doing so, to confirm that removing disks will not impact your production. <<<<
For this purpose, you can use the tool IBM i Performance Data Investigator (PDI) from Navigator for i GUI. Use the performance chart named Disk Throughput Overview for Disk Pools and Physical Disk IO Overview - Basic that look like these:


The point is to take note of disk response time at peak MB/sec. and peak Disk IOPS period(s). If response time does not degrade too much higher than 5 millisec. (I assume HDD rather than SSD here) and/or is not so for a long period of time (say, less than 15 minutes), you are still in good shape. You should look at these charts at peak workload day(s) and you should look at them BEFORE and AFTER you implement PowerHA for i for comparison.
------------------------------
Satid S
Original Message:
Sent: Fri October 04, 2024 11:25 AM
From: Virgile VATIN
Subject: Removing disk units from configuration using SST
Hi John,
For precaution, don't forget to back up your system before removing disk.
I'm not familiar with your production and don't know why 17 disks were needed on the system at startup, but from what I always hear from IBM and for the IBM i in particular the more arms (disks) the better for performance (i/o). So by removing disks from the cluster, you may be increasing processing times. You might want to run a performance analysis before doing so, to confirm that removing disks will not impact your production.
Regards
------------------------------
Virgile VATIN
Original Message:
Sent: Wed October 02, 2024 09:16 AM
From: John Paisley
Subject: Removing disk units from configuration using SST
Hello. Has anyone every used the SST option "Work with removing units from configuration" to remove disks that are in a RAID5 parity set?
We are hoping to make several drives non-configured in order to use them for an IASP. (We are planning to employ PowerHA eventually.)
I'm just wondering whether or not it will work OK and wondering what I can expect.
We are on a Power9 41G at V7R4. We have 17 spinning drives in one RAID set with a hot spare. So the system unit is at capacity as far as disk, and we do not have an additional disk enclosure. We are only running at 41% usage on the system ASP so there is plenty of room to take some drives for the IASP.
Incidentally, I did ask IBM hardware support about this but I am interested to hear of real-world occurrences.
Thanks,
John
------------------------------
John Paisley
------------------------------