Source cleaning cascaded volumes using reference counts

摘要

A system, process, and product for source cleaning cascaded volumes include the ability to store a number of cascaded backup volumes and prepare received writes for storage. The system, process, and product also include the ability to examine a metadata set associated with a volume region for a write to determine whether another volume depends on the region and write the new data to the region if another volume is not dependent. The system, process, and product further include the ability to determine whether a write is for a volume region that is a canonical source and, if so, decrement a reference count in a metadata set associated with the region and establish the volume of the region as a canonical source for a metadata set associated with a new metadata set location identifier.

主权项

1. A method comprising:
storing a primary volume and a plurality of cascaded backup volumes for the primary volume, each volume comprising a plurality of different addressable regions; receiving writes for at least one of the volumes from an external source and preparing the writes for storage; determining whether a write is for a volume region that is a canonical source; if the write is for a region that is not a canonical source:
decrementing a reference count in a metadata set associated with the region,determining a metadata set location identifier,establishing the volume of the region as a canonical source for a metadata set associated with the metadata set location identifier, andestablishing a reference count for the metadata set to indicate that the metadata set is referred to by one volume region; and if the write is for a region that is a canonical source:
examining a metadata set associated with the region to determine whether another volume depends on data in the region, andwriting the new data to the region if another volume does not depend on data in the region.