Policy-Driven Management of Data Sets

Contemporary storage systems separate the management of data from the management of the underlying physical storage media used to store that data. This separation is artificial and increases the overall burden of managing such systems. We propose a new management layer that unifies data and storage management without any loss of control over either the data or the storage. The new management layer consists of three basic entities: data sets, which describe the data managed by the system, policies that specify rules for the management of data sets, and resource pools that represent storage that can be used to implement the policies for the data sets. Using data sets and policies, data administrators are able to directly manage their data, even though that requires them to indirectly manage the underlying storage infrastructure. The storage administrator retains control over what the data administrator can do to the storage. This new management layer provides the infrastructure necessary to build mechanisms that automate or simplify many administrative tasks, including the necessary coordination between the data and storage administrators. We describe Protection Manager, the first tool that uses this management layer to present storage management in the context of data management tasks that the data administrator wants to perform, and evaluate the effectiveness of using this management layer as a way to automate backups.