Dynamic Provisioning: Virtualize it, Snap it, move it, replicate and COW it
by Hu Yoshida on Apr 27, 2008
Unlike other thin provisioning solutions, HDP, or Hitachi Dynamic Provisioning, supports other vendor’s storage that are attached, or virtualized, behind the USP V or VM. Since it is provided as a service on the USP V/VM, it can aggregate thin provisioning with other common storage services provided on this platform. This Services Oriented Approach to HDP not only provides thin provisioning to other vendor’s legacy or lower level storage system through storage virtualization, we can also combine this with other USP V/VM services like snapshot copies, non disruptive moves or migration, distance replication, and Copy on Write and enhance the capabilities of these services.
With HDP we provision 42 MB pages of virtualized storage as a user actually writes to his allocated storage space. When he exceeds his 42 MB page, we give him more pages until he reaches his maximum allocation. If his application enables him to increase his initial allocation request, we can continue to add additional pages.
Since the USP V/VM knows how many pages HDP has provisioned to a volume, when a copy is requested, the USP V/VM only copies the pages that are provisioned and not the entire volume. Without HDP a copy function would have to copy the entire volume, including the allocated unused space, since it would have no idea how much of that volume was actually used. This could provide a tremendous saving in allocated unused storage capacity when one considers how many copies are made for offline back up cycles, replication, development test, data mining, data distribution, etc. Without HDP, all those copies would replicate that allocated unused space, over and over again.
HDP can also be used with dynamic moves for device migration at end of lease, tiered storage for performance tuning or cost reduction, and consolidation following a merger or acquisition. Since only the used pages and not the whole volumes are moved, the moves can be made more efficiently.
This same efficiency applies to synchronous and asynchronous replication and to Copy on Write. Replicating and recoverying active pages rather than allocated volumes will be more efficient and less wasteful of capacity.
This combination of efficencies can only be realized when you take a Services Oriented Approach to thin provisioning, rather than a stand alone approach.
Intersting – you keep asserting that Hitachi Dynamic Provisioning can be used to thin provision external storage. Yet neither Sun nor HP agrees – they both say that Hitachi has not yet released the microcode necessary to enable this capability – and in fact, external storage cannot be selected when creating HDP volumes.
Additionally, TSM does not recognize HDP volumes at all, making it impossible to migrate to/from thin volumes with this utility, even within the array (much less with externally virtualized storage).
What is your explanation for why your claims differ so significantly from what your OEM and Reseller say actually works?