Instead of trying to kill any nodes that should not be participating in the cluster, Data Fencing attempts to restrict access to the necessary data resources so that only the node legitimately running the service gets access to the data (all others being locked out)
Data based fencing gives a much more fine grained approach to data integrity (and one that is much better suited to resource driven clusters). Fencing is most often implemented as a lock placed on the storage itself (via SCSI reservations or via a volume manager using a special data area on storage). This means that if the node can get access to the data, it can also be aware of the locking.
The disadvantage to data based fencing is that it cannot be implemented in the absence of a storage mechanism that supports it (which occurs when the storage is replicated).