What isn't storage virtualization?

Vendors label many products 'virtualization,' but not all of them meet the needs of true storage virtualization

Vendors often use the term "virtualization" to describe myriad products, including global name spaces, virtual storage area networks (VSANs), pooled NAS (network-attached storage), thin-provisioning software, virtual file systems, virtual tape libraries, RAID arrays and disk clusters, and virtualized application and file servers (such as EMC's VMWare). But although these technologies all use some sort of virtualization, they don't actually qualify as storage virtualization.

The term "virtual" applies to many computer technologies; a computer is, after all, a virtual machine that configures itself for each application to accomplish the results of a specific tool. In the storage world, a block is a virtual way to reference data's physical location on a disk's cylinders; a file is a virtual way to reference a series of blocks that make up a set of information; a RAID array is a virtual way to make several physical disks work as if they were one; and so on. The goal for all forms of virtualization is to abstract the data from the hardware that contains it, so as data flows the services and applications that use it don't need to be reconfigured.

True storage virtualization, however, is the ability to virtualize arrays, servers, and other storage media from a variety of vendors, with a variety of operational characteristics -- speed, firmware, capacity, and interface -- in a networked environment. You could think of it as a way to create a network-level RAID from heterogeneous storage devices, so the entire enterprise storage environment becomes one system.

Technologies such as VSANs (essentially, logical partitions on a SAN), pooled NAS, and virtual file systems don't provide that, but they can be quite useful in a storage environment. For example, file-oriented virtualization techniques, such as global name spaces and virtual file systems, let users and applications access files easily no matter where they are stored in the enterprise. Essentially, they create a virtual volume of all files by using metadata to map physical data locations to logical, virtual locations, and then present them to the user as a single network "drive." Though not as sophisticated as storage virtualization products from the likes of EMC, IBM, and NetApp, technologies like these often serve as building blocks on the way to true storage virtualization, says Gartner research director Stan Zaffos.