DB and Log

PREDATAR Control23

Will it work? Yes
Will it perform? No

Unless it's a small test server, you will be disappointed with performance, spinning disks can't handle the IOPS requirements. More info:
https://www.ibm.com/support/knowledgecenter/en/SSEQVQ_8.1.8/perf/r_chklist_srvdbdsk.html

You want a minimum of 4 SSD LUNs for the DB, and 1 SSD LUN for the active log. If the DB is greater than 2TB, you want to increase that to 8 SSD LUNs for the DB. And if greater than 4 TB, 12 SSD LUNs.
 
PREDATAR Control23

Will it work? Yes
Will it perform? No

Unless it's a small test server, you will be disappointed with performance, spinning disks can't handle the IOPS requirements. More info:
https://www.ibm.com/support/knowledgecenter/en/SSEQVQ_8.1.8/perf/r_chklist_srvdbdsk.html

You want a minimum of 4 SSD LUNs for the DB, and 1 SSD LUN for the active log. If the DB is greater than 2TB, you want to increase that to 8 SSD LUNs for the DB. And if greater than 4 TB, 12 SSD LUNs.

I know about IBM's recommendation but I beg to disagree.

All of my TSM servers are attached to Netapp storage with spinning disks on these when we started. Some TSM servers have a load of 1800+ nodes with 2 or more TiB of DB requirement. Backups do finish within the required 6 hour window with DB backup finishing in 30 mnutes or less.

The point to all of this is how the storage handles traffic - is there buffering? is QOS in place, etc.

Now that we have slowly moved to all SSD, I don't notice any difference. Still completes the job at the same time as mentioned above.
 
Last edited:
Top