Hardware snapshots

1. A mirrored pair is established between a volume and its user-controlled mirror. The data on the mirror remains synchronized with the data on the standard volume.
2. When the snapshot utility job is submitted, the utility registers the data sets with XBM and snapshot processing starts.
3. The snapshot utility issues a quiesce or otherwise establishes the necessary point of consistency.
4. The snapshot utility indicates to XBM that the snapshot should be started:
   • For a volume-level snapshot, the user-controlled mirror is separated from the standard volume where the registered data sets reside. This separation splits the link between the two volumes, leaving a point-in-time copy of the data on the mirror. Updates to the standard volume can continue.
   • For a data-set-level snapshot, XBM issues the appropriate request to have the data set replicated, or snapped, within the intelligent storage control unit. A hardware-vendor-specific API is called to create a point-in-time copy of the data set. Updates to the registered data sets can continue.
5. When the snapshot utility begins processing data, XBM provides the snapshot utility with access to the preimage data on the mirror or in the replicated data set. Doing so allows the snapshot to process the data as it was before any concurrent update activity.
6. When snapshot processing is completed, XBM initiates reconnection of the volumes or deletes the temporary data sets.

If the volumes or data sets cannot be separated when the snapshot process begins (for example, if the volumes are busy reconnecting from an earlier separation), XBM can seamlessly fall back to a traditional software snapshot by using the XBM cache to provide preimages to the utility.