questions regarding wasted space and tunesys

[Fastjack]

Hi,

I'm currently working with a V800 using Thin Provisioning and Adaptive Optimization with 3 tiers (NL, FC and SSDs).
Some of our production-volumes have been operational for a while now and looking at a "showvv -s" I have noticed what I consider lots of wasted space. In the category Usr-space some volumes show a large difference between reserved and used space. While each volume uses around 3TB space the reserved space is up to 2TB larger than that. Over all volumes the difference is around 15%.
At first I hoped that over time the system would clean this up by itself but more than a week has gone by without change.
How do I get the system to free up this wasted space?

One thing I tried to resolve the above problem was to just start a tunesys (no idea whether that would have helped). But I soon found that tunesys in turn started a number of "tunetpvv snp_cpg ...".
At this point I canceled the tunesys as I was afraid that the tunetpvvs (or rather those that were still to come) would conflict with the Adaptive Optimization by moving all the volume-data back to their default CPGs.
So my second question is: Would a tunesys indeed mess up the storage tiering?

Thanks in advance

[Richard Siemers]

What I believe (hypothesis) is happening with that 15% is that to prevent a yo-yo effect of cycling back and forth between allocated and unallocated.. and probable unnecessary churning of work... volumes that have grown and shrunk back down, use some algorithm so that the system will not reclaim 100% of that space, and instead keep a buffer as "reserved" for immediate growth. I believe this helps by avoiding excessively fragmenting the data, and keeping growth in large chunks of contiguous blocks.

[Christian]

What you are saying Richard is most likely correct.

I'm doing a sdelete on a few 4TB TPPV that was 100% filled on the 3PAR and i Windows they had 2TB of data free

I'm seeing that after a sdelete my revered space is about 3.6TB but it slowly going down towards the Used Usr Size.

So your assumption that its due to preventing a yo-yo effect sounds correct to what I'm seeing atm.

But it seems it only on 3.1.1 that this is happening , on 2.3.1 I have never encounter this problem when running sdelete and compact CPG.

[Fastjack]

If the system always left a bit of buffer-space I would be fine with it. But in some cases the difference is quite extreme. One of our volumes has 3.5TB reserved user-space with only 14GB actually used (as seen with showvv -s). This has not changed for weeks now since the volume is currently not actively used. In System Reporter I can also see that Adaptive Optimization has moved all the volume's 3.5TB to NL-discs by now. So this "buffer-space" actually occupies space on the discs.

The compactcpg tasks the system runs regularly don't seem to clean up this sort of wasted space which makes sense, since the space is wasted on the volume-level. So what I am looking for is some command that optimizes/compacts on the volume-level.

BTW: We're running version 3.1.1

[brettcarr]

Has anybody worked out a definitive answer on this, we are seeing the same issue, I currently have over 6tb wasted (Stuck in reserved within volumes) accross two different arrays. I'd really like to get this space back.

Thanks

Brett