Storage Tank is IBM's storage virtualization architecture that uses a distributed object model to provide storage to clients on the system. It is a SAN-based distributed object storage system aimed at use in heterogeneous environments. Storage Tank increases performance by separating the storage metadata from the data stream and allows clients to access data directly. This in itself reduces overhead, thus improving performance. But the key to Storage Tank is IBM's Installable File System (IFS).
Unlike most storage virtualization approaches, Storage Tank uses its own file system, the Installable File System (IFS). IFS is installed on each client and handles the interaction between the client and the data. In the best traditions of virtualization, none of this is visible to the client since the IFS presents the data in the client's native format.
Since Storage Tank relies on its own file system, devices on the system must be capable of using that file system. In the case of clients and servers this is a simple matter of installing the lightweight IFS client to translate between the native file system and IFS. However in a modern IT system not everything uses the file system. For reasons of performance some applications, notably databases, go around the file system. Making these applications Storage Tank compatible requires using a special API on the application.
An overview of Storage Tank and how it works can be found at: www.usenix.org/publications/library/proceedings/fast02/wips/pease.pdf.
Rick Cook has been writing about mass storage since the days when the term meant an 80K floppy disk. The computers he learned on used ferrite cores and magnetic drums. For the last twenty years he has been a freelance writer specializing in storage and other computer issues.