Frame swap challenge when using Delayed Premigration. I call it "Autonomic prepare of frame swap when using Delay Premigration”

Below idea is related to frame swap from old TS7700T subsystem to new TS7700T. Meaning virtual tape having a backend Tape Library(ie TS4500) and backend Pdrives/Pvols.

When preparing frame swap, due to hardware renewal, you need to be well prepared. Know the current setting as well as having some knowledge on how IBM can conduct the frame swap.

We use Delay Premigration for Two Cluster setups. It is well functioning, and it saves a lot of PDRIVE/PVOL activity as well as a lot of backend Reclaim activity. We save data as two copies in cache only for at least 200 days and the many TBs never touches the backend devices. A fine functionality.

But!

A frame swap requires that all data in cache have a copy on backend medias. Otherwise, the frame swap activity will take several days.

Days where the subsystem are in maintmode/offline for z/OS and where lvols aren’t mirrored. We/IBM need to be able to conduct such an activity in hours instead of in days.

Today I prepare for such a frame swap. Such activity completely manually and occupies a lot of time.

How to conduct it is related to number of Features installed called “x TB Active Premigration Queue”. Could be FC 5274(1 TB increment) or FC 5259(5TB increments) and many other factors.

You need to prepare in microsteps. In my case reduce in parts of 5-8 TB and have a lot of manual controls.

I have to prepare for more than 400TB per subsystem and I have six of them!

Worst case if you conduct it wrongly, I will get serious throttling of all I/O to tape. Nobody want to see that!

We need IBM to support such activity with an automated function without causing Throttle.

We could call it “Autonomic prepare of frame swap when using Delay Premigration”