File virtualization is the creation of an abstraction layer between file servers and the clients that access those file servers. Once deployed, the file virtualization layer manages files and file systems across servers, allowing administrators to present clients with one logical file mount for all servers. The file servers continue to host file data and metadata.
While this arrangement may seem to add IT overhead unnecessarily, file virtualization provides several key advantages, including a global namespace to index files on network file servers. In addition, this virtual file storage consolidation allows excess storage capacity to be shared between file servers. Data migrations can be conducted between file servers transparently to end users and applications, which is ideal in a tiered storage infrastructure. In short, file virtualization allows companies to access isolated ("stranded") storage capacity on network file servers and perform seamless file migrations among them.
File virtualization may be deployed as an appliance or as an off-the-shelf server running file virtualization software. The choice is generally based on cost and the level of management and disruption involved. Appliances, the most common deployment option, are available in four different architectures; out-of-band, in-band, a combination of out-the two, and split-path.
Not all file virtualization deployments are successful over the long term, and some organizations may need to roll back (remove) their deployment. This can be a highly disruptive process to file servers and NAS (network-attached storage) platforms. In extreme cases, a rollback may require an organization to unload data, remove the file virtualization layer, and then reformat and reload all of the data from scratch. Vendors can often help to identify potential rollback problems and offer advice to mitigate the disruption. Users often test their rollback procedures before general deployment.
File virtualization can be limited by scalability, which can involve the number of file systems, files, servers or I/O (input/output) performance. The file virtualization platform must also be interoperable with the current infrastructure so that it can work with existing storage systems and switches. To prevent potential problems, a file virtualization platform should always be tested for adequate scalability and interoperability.