This operation is a Vault extension to the Amazon S3 API. The operation enables an existing bucket to be configured as a "virtual bucket". A virtual bucket is a bucket of unlimited size for which object storage spans multiple Vault service regions. The virtual bucket feature is applicable only for a multi-region Vault deployment. Its purpose is to accommodate buckets with larger capacities than service providers may wish to allow in a single region.

When a bucket is virtualized, objects PUT into that bucket may be stored in the same region in which the bucket was created, or in any other region in the Vault service deployment. The system’s determination as to which region to store the object into is based on a consistent hashing scheme. Each region is assigned a token (forming a token ring). When the S3 Server in the region in which the bucket was created subsequently receives PUT Object requests for the bucket, it assigns each object a token (derived from a hash of the bucket name and object name) and then transmits the object to the region whose token range encompasses the object’s token (which may be the local region or any of the other regions).

Subsequent requests to retrieve (or delete) an object stored in a virtual bucket are handled by the S3 service in the region in which the bucket was originally created. The S3 service routes the request to the correct region. This action is invisible to the client. After the correct region returns a response to the S3 service in the region that’s handling the request, the S3 service returns a response to the client.

Regardless of which regions a virtualized bucket’s objects are stored in, all usage data for the bucket (for purposes of QoS enforcement and billing) is tracked at the region in which the bucket was originally created.