Novel experimental approaches often necessitate the creation of databases that exist outside the scope of public sequence repositories. Typically, these databases are hosted remotely and users interact with them via a web-based interface.

Since these databases are often used as experimental reagents in their own right in computational studies, users should be provided a mechanism to easily cite specific objects and database views, usually as a URL. But rapidly evolving content can make stable referential access to previous versions of the database onerous and costly, requiring multiple versions of the resource and oftentimes multiple servers. Virtualization -- or the ability to run one operating system entirely within another -- offers an easy solution to these challenges.

Here we present our experiences using virtualization to create referential "freezes" of a large genomic annotation database. These "virtual machines" contain a guest operating system, all components of the database, and the web-based interface to the database. Creation of these virtual machines can be scripted, providing a convenient and consistent mechanism forversioning the resource.

Virtual machines can also be used to distribute a web-based resource. This gives end users the option of using the resource locally on their platform of choice without requiring any expertise in installing the database or software dependencies. Local installations using a virtual machine offer other benefits including the opportunity to integrate private data within the framework of a highly developed public informatics resource, flexible data mining, and facile mirroring.