Which feature is used to roll back a corrupted HDFS instance to a previously known good point in time?
partitioning
high availability
replication
snapshot