It's almost a given that storage virtualization will involve some form of thin provisioning technology to make the most efficient use of available resources. Still up in the air, though, is which of the many flavors of TP is most effective and what kind of data policy framework needs to be in place.
Thin provisioning is the ability to assign storage on an as-needed basis, rather than the usual practice of over-provisioning storage to suit anticipated requirements. TP is likely to become a crucial component of virtual SAN technology that otherwise would face a severe capacity crunch in the face of numerous virtual machines and desktops.
One of the better round-ups on thin provisioning issues is this article by Rich Friedman on Search Storage. He points out some of the differences between some of the major TP implementations, like their maximum supported pool size and how "granular" your data chunks need to be. As for policy management, they will likely differ from primary to mirror sites, and even across various storage tiers.
Most of the major storage vendors have already embraced thin provisioning in their top platforms. IBM was one of the last to join the party, coming out this week with an updated version of its SAN Volume Controller (SVC) software for the System Storage family. The company says its Space-Efficient Virtual Disks feature works across heterogeneous environments, allowing it to be quickly integrated into legacy systems.
EMC looks like it's taking a slightly different approach, pitching its solution as "virtual provisioning," slated to become part of the Clariion, Celerra, Symmetrix and Connetrix families within the next year, followed by a steady rollout across backup and recovery systems. The company says its approach provides for greater automation of the provisioning process by constantly monitoring available resources and balancing them against application demands.
Not everyone is impressed with the virtual provisioning strategy, however. Barry A. Burke, aka The Storage Anarchist, claims it will lead to a "virtual provisioning catch 22," or rather, a series of catch 22s that produce even greater headaches for administrators than simple provisioning ever did. He says key problems include runaway applications exceeding their allocations, deleted memory failing to return to the storage pool, and I/O bottlenecks caused by too many users trying to access the same disks.
It would be tempting to use a catchy phrase like "Thin is In" right about now, and certainly the vendors seem to have caught the bug. But thin provisioning isn't one of those technologies that you just plug in and watch go to work. Configured correctly, it can be a real bonus. Otherwise, it might leave you longing for the good ol' days.