October 06, 2008, 4:34 PM — Survey after survey indicates that interest in and adoption of storage virtualization is increasing. But I cringe a bit when I see those surveys, because it's often impossible to determine what the numbers really tell us. How does each respondent define storage virtualization?
Unlike server virtualization, which represents a clearly understood product category, storage virtualization has come to represent an assortment of technologies and approaches. Given vendors' eagerness to attach their products in any way, shape or form to the "V" word, more often than not, the term storage virtualization requires more explanation.
Virtualization, after all, is a technique that over the years has been employed in almost every facet of storage technology: disk, controller, array, SAN, LAN and even host (think logical volume manager). At the risk of oversimplification, I'd suggest that there are three broad categories: inside the box, outside the box and tape emulation. Each of these offers distinct benefits as well as some commonalities.
Inside-the-box virtualization. All storage virtualization approaches involve pooling some quantity of physical storage and creating logical abstractions of subsets of that storage. "Inside the box" refers to storage systems that make use of virtualization internally. You could say that this would include every storage array sold today, and, in a sense, you'd be right. It's the degree of abstraction that differentiates a "traditional" array from a virtualized one. A virtualized array aggregates large numbers of disks, often in a gridlike manner, and then creates and presents virtualized volumes from those disks. Thus, I/O operations typically span a larger number of spindles than in a traditional array.
What are the benefits? One would be ease of management -- creating, presenting, expanding, shrinking, copying and deleting logical storage in these systems involves fewer steps than with traditional arrays. Another can be performance -- a common bottleneck is the I/O capability of a small set of disks, and by striping widely across many spindles, virtualized arrays can overcome this limitation.
Outside-the-box virtualization. This technology was designed to provide a virtualized layer of common services that can be applied across a number of potentially heterogeneous arrays. To many, this represents the true intent of the term storage virtualization -- the ability to aggregate storage pools beyond the confines of a single array. This approach leverages various combinations of intelligent switches and appliances to provide volume management, snapshot, cloning and replication capabilities.
Moving functionality to the network layer and outside of the storage array yields the key benefit of common management of heterogeneous physical storage devices. Originally positioned as a means of commoditizing storage systems, the primary (but not sole) applications include data migration and replication. Clustered network-attached storage grids, a form of outside-the-box virtualization, also provide uniform capacity and performance scalability.
Tape emulation. As represented by virtual tape libraries, tape emulation is probably the nearest thing that the storage world has to a "pure" virtualization product category; it involves abstracting disk storage to appear as tape storage. Backup is a perennial problem for most organizations, and introducing disk to the backup process in the form of a VTL is a relatively easy way to address many of the performance and management issues associated with tape backup.
As with any generalization, there are bound to be exceptions. One example is Hitachi Data Systems' Universal Storage Platform family, where outside-the-box features are found within the array controller rather than at the network layer.
The movement to virtualized storage is inevitable. Unfortunately, the options can be confusing. Rather than focusing on virtualization per se, understand and prioritize requirements and the functions needed to meet them.