We are in the process of implementing a new VMAX array and adding mostly new servers and migrating a few existing applications.
One big procedural difference doing mostly thin provisioning is the ability to reclaim space from active servers/applications/etc.
Other than post or during migration what is everyone else doing as far as scheduling reclamations on active applications/volumes? Doing it based on a schedule, major application changes, etc.
Various platforms in effect including VMWare, Windows, and various UNIX variants.
We have a decent sized mixed env (Windows, VMware, AIX, Linux, iSeries, zSeries) and during the initial few months played around with scheduling regular reclaims against the thin pools. We saw next to no gains as far as space reclamation goes, which makes sense given how most filesystems don't zero out files upon deletion. It will not deallocate all zeros either, you need to have a contiguos 768k of zeros within the individual extent boundary for that extent to be freed.
There is a good whitepaper on powerlink on zero reclaim, where the biggest gains will come from is the new UNMAP SCSI command, however there isn't a great deal of support for it so far in the opensystems world. EMC built a tool to run against NTFS filesystems that issues UNMAPs against the free space, and I believe VAAI on vmware also provides this in a more automated fashion.
The document isn't clear on whether a reclaim job is needed after UNMAP commands are sent, or if the extents are deallocated inline.
Link to the whitepaper:
if that doesn't work you can navigate to it:
"White Paper: EMC Symmetrix VMAX Virtual Provisioning Space Reclamation and Application Considerations — Applied Technology"
Not really looking for explanation of how it works - rather curious as to how other people are using it.
After migrations is an obvious time to run a reclaim but how are people using it in the real world. When do you say "it's time to run a reclaim" for different types of databases, environments, etc.
Currently with the way OS support this it seems to require some extra work on the system administrators side to really get any benefit from this reclaim process so maybe in practice this is only useful during migrations/cleanup.
At the start of my last reply...
"We have a decent sized mixed env (Windows, VMware, AIX, Linux, iSeries, zSeries) and during the initial few months played around with scheduling regular reclaims against the thin pools. We saw next to no gains as far as space reclamation goes"
We use it after OR migrations as you stated (although no longer necessary if on 5875), and never any other time.